interaction_id stringlengths 9 9 | session_id stringlengths 6 6 | user_id stringclasses 438 values | timestamp_utc stringdate 2025-02-01 02:22:18 2025-04-30 22:23:19 | split stringclasses 3 values | channel stringclasses 5 values | use_case stringclasses 6 values | country_code stringclasses 10 values | region stringclasses 3 values | account_tier stringclasses 3 values | segment stringclasses 3 values | instruction_template stringclasses 6 values | model_provider stringclasses 6 values | model_name stringclasses 8 values | temperature float64 0 1.2 | top_p float64 0.71 1 | max_tokens int64 256 2.05k | prompt_tokens int64 13 1.91k | completion_tokens int64 5 1.12k | total_tokens int64 29 2.75k | latency_ms int64 156 5.62k | cost_usd float64 0 1.05 | response_quality_score float64 0.09 1 | is_failure bool 2 classes | failure_type stringclasses 7 values | hallucination_flag bool 2 classes | toxicity_flag bool 2 classes | safety_block_flag bool 2 classes | formatting_error_flag bool 2 classes | tool_error_flag bool 2 classes | latency_timeout_flag bool 2 classes | tool_calls_count int64 0 3 | tools_used stringclasses 65 values | user_feedback_score float64 -2 2 ⌀ | user_feedback_label stringclasses 4 values | user_reported_issue bool 2 classes | retry_index int64 0 3 | repair_strategy stringclasses 7 values | final_resolution_state stringclasses 3 values | business_impact_tag stringclasses 4 values | request_text stringlengths 63 301 | response_text_snippet stringlengths 180 263 | tokens_per_second float64 27.4 7.66k | date_utc stringdate 2025-02-01 00:00:00 2025-04-30 00:00:00 | hour_of_day_utc int64 0 23 | day_of_week int64 0 6 | is_weekend bool 2 classes | is_peak_hour bool 2 classes | prompt_to_completion_ratio float64 0.02 208 | request_text_template stringclasses 140 values | instruction_text stringclasses 6 values | user_feedback_observed bool 2 classes | prompt_id stringclasses 36 values | response_text_snippet_len_chars int64 180 263 | response_text_snippet_is_truncated bool 1 class |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
INT000101 | S00716 | U00190 | 2025-03-13T00:45:56Z | train | internal_tool | customer_support | AU | APAC | pro | enterprise_team | internal_policy_qa | anthropic | claude-3.5-sonnet | 0.1 | 0.882 | 256 | 1,069 | 256 | 1,325 | 1,409 | 0.424 | 0.662 | false | none | false | false | false | false | false | false | 1 | knowledge_lookup | 1 | positive | false | 0 | none | auto_resolved | low | Context for you: I need a short, friendly reply to a customer who is asking about the status of their refund. I'd like a practical, engineering-focused answer. | We can make this robust with a few changes. Goal: handle slow responses reliably, with guardrails for formatting and failure handling. Plan: - Use schema validation (e.g., Pydantic) to prevent bad inputs from propagating. - Add telemetry fields... | 940.383251 | 2025-03-13 | 0 | 3 | false | false | 4.1758 | I need a short, friendly reply to a customer who is asking about the status of their refund. | You are a helpful customer support assistant for a SaaS product. Answer clearly, empathetically, and step-by-step while keeping the response concise and practical. | true | 03bf57d7adc83026d525c8192379aa203b71e1eb | 247 | true |
INT000102 | S01257 | U00134 | 2025-04-10T00:40:04Z | train | api | customer_support | IN | APAC | free | individual | support_strict | anthropic | claude-3.5-sonnet | 0.604 | 0.925 | 1,536 | 198 | 286 | 484 | 1,250 | 0.15488 | 0.858 | false | none | false | false | false | false | false | false | 0 | none | null | no_feedback | false | 1 | regenerate_with_lower_temp | escalated_to_human | high | Context for you: Follow-up: the previous reply did not fully solve the issue. Our customer is getting an error when trying to log in to their account. Can you help me debug this? Can you walk me through a good solution? | Let's break this down into actionable steps. Goal: handle billing dispute reliably, with guardrails for formatting and failure handling. Plan: - Add telemetry fields for latency/tokens/cost so you can debug regressions quickly. - Use schema... | 387.2 | 2025-04-10 | 0 | 3 | false | false | 0.6923 | Follow-up: the previous reply did not fully solve the issue. Our customer is getting an error when trying to log in to their account. Can you help me debug this? | You are a helpful customer support assistant for a SaaS product. Answer clearly, empathetically, and step-by-step while keeping the response concise and practical. | false | d029d0919f8e306d09ca7001d500c1e6f1c81007 | 243 | true |
INT000103 | S00934 | U00302 | 2025-02-06T15:44:28Z | train | api | coding_assistant | EG | EMEA | free | team | creative_writer | openai | gpt-4o-mini | 0.219 | 0.882 | 512 | 975 | 320 | 1,295 | 897 | 0.02588 | 0.676 | true | hallucination | true | false | false | false | false | false | 1 | workflow_orchestrator | null | no_feedback | false | 2 | decompose_task | escalated_to_human | medium | Follow-up: the previous reply did not fully solve the issue. I'm getting a Java error: `ValueError`. Can you explain what it means and how to fix it? Please explain step by step. | Let's turn this into a reliable workflow. Signal: hallucination while working on rate limiting. Quick diagnosis: Refuse when confidence is low; avoid confident guesses. Recommended actions: -... | 1,443.701226 | 2025-02-06 | 15 | 3 | false | true | 3.0469 | Follow-up: the previous reply did not fully solve the issue. I'm getting a Java error: `ValueError`. Can you explain what it means and how to fix it? | You are a senior software engineer helping users with code, debugging, and best practices. Provide clear explanations, structured steps, and high-quality code examples when relevant. | false | 83c939e9b891e505ac06548e158556e919317a64 | 194 | true |
INT000104 | S01168 | U00115 | 2025-02-15T15:05:17Z | train | web_app | coding_assistant | US | AMER | free | individual | creative_writer | meta | llama-3.1-70b | 0.186 | 0.897 | 2,048 | 714 | 733 | 1,447 | 1,699 | 0.2894 | 0.537 | true | latency_timeout | false | false | false | false | false | true | 0 | none | -2 | negative | false | 0 | enable_tools | escalated_to_human | low | Context for you: Can you refactor this function to be more readable and efficient while keeping the same behaviour? | Here's what I'd implement in production. Let's stabilize it. Signal: latency_timeout while working on FastAPI endpoint design. Quick diagnosis: Reduce payload size and enable caching for repeated queries. Recommended actions: - Implement... | 851.677457 | 2025-02-15 | 15 | 5 | true | true | 0.9741 | Can you refactor this function to be more readable and efficient while keeping the same behaviour? | You are a senior software engineer helping users with code, debugging, and best practices. Provide clear explanations, structured steps, and high-quality code examples when relevant. | true | 83c939e9b891e505ac06548e158556e919317a64 | 240 | true |
INT000105 | S01021 | U00235 | 2025-02-26T17:01:51Z | train | slack | customer_support | BR | AMER | free | individual | general_assistant | openai | gpt-4o-mini | 0.553 | 0.903 | 2,048 | 761 | 839 | 1,600 | 1,133 | 0.032 | 0.748 | false | none | false | false | false | false | false | false | 0 | none | null | no_feedback | false | 0 | none | escalated_to_human | medium | Context for you: I need a short, friendly reply to a customer who is asking about the status of their refund. Please explain step by step. | Here's the clean way to approach it. Goal: handle billing dispute reliably, with guardrails for formatting and failure handling. Plan: - Prefer simple baselines first; then iterate based on measured failure slices. - Use schema... | 1,412.180053 | 2025-02-26 | 17 | 2 | false | true | 0.907 | I need a short, friendly reply to a customer who is asking about the status of their refund. | You are a helpful customer support assistant for a SaaS product. Answer clearly, empathetically, and step-by-step while keeping the response concise and practical. | false | a52fb80193c7c0769c0836852624f4b025a04d2a | 230 | true |
INT000106 | S00203 | U00108 | 2025-02-10T03:29:56Z | train | web_app | internal_qa | AU | APAC | pro | enterprise_team | analytics_helper | meta | llama-3.1-70b | 0.595 | 0.898 | 768 | 66 | 419 | 485 | 559 | 0.097 | 0.682 | false | none | false | false | false | false | false | false | 0 | none | 1 | positive | false | 0 | none | auto_resolved | low | Can you help me debug this? Explain our refund policy in simple language that a new support agent can remember. A short, clear answer would be ideal. | Here's the clean way to approach it. Goal: handle tool failures reliably, with guardrails for formatting and failure handling. Plan: - Add telemetry fields for latency/tokens/cost so you can debug regressions quickly. - Start... | 867.620751 | 2025-02-10 | 3 | 0 | false | false | 0.1575 | Explain our refund policy in simple language that a new support agent can remember. | You are an internal knowledge assistant helping teammates navigate documentation, policies, and internal tools. Be precise, concise, and point to relevant information and next steps. | true | d33eca392020f2e6e14ec7fe3091a25b46d6bae1 | 228 | true |
INT000107 | S01313 | U00060 | 2025-03-04T08:38:57Z | train | web_app | brainstorming | FR | EMEA | free | individual | support_strict | openai | gpt-4.1 | 0.476 | 0.97 | 512 | 101 | 452 | 553 | 367 | 0.2212 | 0.6 | false | none | false | false | false | false | false | false | 0 | none | 1 | positive | false | 0 | none | auto_resolved | low | Can you help me debug this? Give me ten creative ideas for a YouTube series about learning data science from scratch. I'd like a practical, engineering-focused answer. | Let's make the behavior deterministic and observable. Goal: handle customer support automation reliably, with guardrails for formatting and failure handling. Plan: - Use schema validation (e.g., Pydantic) to prevent bad inputs from propagating.... | 1,506.811989 | 2025-03-04 | 8 | 1 | false | false | 0.2235 | Give me ten creative ideas for a YouTube series about learning data science from scratch. | You are a creative assistant helping users explore ideas and options. Suggest multiple alternatives, structure them clearly, and stay practical. | true | e3da6c15e87a6fab71f2d2b735988233f5d1f30b | 247 | true |
INT000108 | S00979 | U00223 | 2025-02-23T14:53:46Z | val | web_app | content_writing | AE | EMEA | free | team | support_strict | openai | gpt-4.1 | 0.96 | 0.818 | 1,024 | 479 | 533 | 1,012 | 994 | 0.4048 | 0.748 | false | none | false | false | false | false | false | false | 0 | none | null | no_feedback | false | 0 | none | auto_resolved | high | Turn these bullet points into a short, engaging blog introduction about learning machine learning. I'd like a practical, engineering-focused answer. | Here's a clear, step-by-step answer. Goal: handle observability metrics reliably, with guardrails for formatting and failure handling. Plan: - Use schema validation (e.g., Pydantic) to prevent bad inputs from propagating. - Add... | 1,018.108652 | 2025-02-23 | 14 | 6 | true | true | 0.8987 | Turn these bullet points into a short, engaging blog introduction about learning machine learning. | You are a writing assistant helping users craft clear, engaging, and well-structured content. Keep the tone natural, readable, and aligned with the user's intent. | false | 75b7366387b6cf4f5631ec9dd298a359e5bb9c1d | 230 | true |
INT000109 | S01065 | U00046 | 2025-04-22T03:36:08Z | train | slack | coding_assistant | IN | APAC | free | team | analytics_helper | local | custom-local-8b | 0 | 0.939 | 1,536 | 297 | 237 | 534 | 928 | 0.00212 | 0.656 | true | formatting_error | false | false | false | true | false | false | 1 | workflow_orchestrator | null | no_feedback | true | 0 | add_constraints_to_prompt | auto_resolved | high | Quick question: I need help understanding why this asynchronous code is not awaiting correctly. Explain with a small code snippet. Please explain step by step. | Let's break this down into actionable steps. Signal: formatting_error while working on caching strategy. Quick diagnosis: Use constrained generation (schema-first prompt) and add unit tests for... | 575.431034 | 2025-04-22 | 3 | 1 | false | false | 1.2532 | I need help understanding why this asynchronous code is not awaiting correctly. Explain with a small code snippet. | You are a senior software engineer helping users with code, debugging, and best practices. Provide clear explanations, structured steps, and high-quality code examples when relevant. | false | 54f94ac4f6df7ccc4808605ae4fb839b7a1fbfb2 | 196 | true |
INT000110 | S00046 | U00438 | 2025-02-08T09:51:38Z | train | web_app | content_writing | GB | EMEA | enterprise | individual | general_assistant | openai | gpt-4o-mini | 0.873 | 0.941 | 1,024 | 1,234 | 339 | 1,573 | 1,233 | 0.03144 | 1 | false | none | false | false | false | false | false | false | 0 | none | 1 | positive | false | 0 | none | auto_resolved | low | Quick question: Write a concise product description for a new wireless mouse aimed at software developers. | Here's what I'd implement in production. Goal: handle LLM reliability reliably, with guardrails for formatting and failure handling. Plan: - Add telemetry fields for latency/tokens/cost so you can debug regressions quickly. - Start by writing a... | 1,275.750203 | 2025-02-08 | 9 | 5 | true | true | 3.6401 | Write a concise product description for a new wireless mouse aimed at software developers. | You are a writing assistant helping users craft clear, engaging, and well-structured content. Keep the tone natural, readable, and aligned with the user's intent. | true | a1c74d04f735d164fdc621e3ce3d5bfea1026243 | 247 | true |
INT000111 | S00564 | U00209 | 2025-02-26T07:16:34Z | train | web_app | customer_support | IN | APAC | pro | individual | support_strict | anthropic | claude-3.5-sonnet | 0.729 | 0.898 | 256 | 1,435 | 256 | 1,691 | 1,038 | 0.54112 | 0.94 | false | none | false | false | false | false | false | false | 0 | none | 1 | positive | false | 0 | none | auto_resolved | low | Hi, I'm having an issue with my billing. The invoice total doesn't match what I expected. Can you check it? | Here's what I'd implement in production. Goal: handle account verification reliably, with guardrails for formatting and failure handling. Plan: - Use schema validation (e.g., Pydantic) to prevent bad inputs from propagating. - Start by writing a... | 1,629.094412 | 2025-02-26 | 7 | 2 | false | false | 5.6055 | Hi, I'm having an issue with my billing. The invoice total doesn't match what I expected. Can you check it? | You are a helpful customer support assistant for a SaaS product. Answer clearly, empathetically, and step-by-step while keeping the response concise and practical. | true | d029d0919f8e306d09ca7001d500c1e6f1c81007 | 248 | true |
INT000112 | S00717 | U00207 | 2025-04-13T08:38:31Z | train | web_app | coding_assistant | EG | EMEA | free | individual | creative_writer | openai | gpt-4o-mini | 0.324 | 0.916 | 1,024 | 133 | 400 | 533 | 3,177 | 0.01068 | 0.667 | true | tool_error | false | false | false | false | true | false | 1 | tool_invocation_failed | -2 | negative | false | 0 | decompose_task | escalated_to_human | critical | I'm stuck on this and need some guidance: Write a small example that shows how to handle exceptions properly in Java. A short, clear answer would be ideal. | We can make this robust with a few changes. Let's stabilize it. Signal: tool_error while working on API retry logic. Quick diagnosis: Validate tool arguments before calling the tool; reject invalid calls early. Recommended actions: - Log... | 167.768335 | 2025-04-13 | 8 | 6 | true | false | 0.3325 | Write a small example that shows how to handle exceptions properly in Java. | You are a senior software engineer helping users with code, debugging, and best practices. Provide clear explanations, structured steps, and high-quality code examples when relevant. | true | 83c939e9b891e505ac06548e158556e919317a64 | 240 | true |
INT000113 | S00440 | U00403 | 2025-02-20T09:54:54Z | train | api | coding_assistant | DE | EMEA | free | individual | creative_writer | meta | llama-3.1-70b | 0.297 | 0.922 | 1,024 | 1,043 | 560 | 1,603 | 881 | 0.3206 | 0.98 | false | none | false | false | false | false | false | false | 3 | code_run|email_send|vector_search | 1 | positive | false | 1 | add_constraints_to_prompt | escalated_to_human | low | I'm working on something and need help: Follow-up: the previous reply did not fully solve the issue. Can you refactor this function to be more readable and efficient while keeping the same behaviour? I'd like a practical, engineering-focused answer. | Here's the clean way to approach it. Goal: handle caching strategy reliably, with guardrails for formatting and failure handling. Plan: - Start by writing a small function boundary for caching strategy and add tests before optimizing. - Prefer... | 1,819.523269 | 2025-02-20 | 9 | 3 | false | true | 1.8625 | Follow-up: the previous reply did not fully solve the issue. Can you refactor this function to be more readable and efficient while keeping the same behaviour? | You are a senior software engineer helping users with code, debugging, and best practices. Provide clear explanations, structured steps, and high-quality code examples when relevant. | true | 83c939e9b891e505ac06548e158556e919317a64 | 246 | true |
INT000114 | S00773 | U00410 | 2025-02-10T16:52:38Z | train | mobile_app | internal_qa | AE | EMEA | free | enterprise_team | analytics_helper | anthropic | claude-3.5-sonnet | 0.237 | 0.827 | 768 | 383 | 236 | 619 | 1,144 | 0.19808 | 0.638 | true | latency_timeout | false | false | false | false | false | true | 0 | none | null | no_feedback | true | 3 | decompose_task | escalated_to_human | high | Context for you: Follow-up: the previous reply did not fully solve the issue. Summarise this internal security policy into three bullet points for non-technical employees. A short, clear answer would be ideal. | Let's make the behavior deterministic and observable. Signal: latency_timeout while working on token spikes. Quick diagnosis: Reduce payload size and enable caching for repeated queries. Recommended actions: -... | 541.083916 | 2025-02-10 | 16 | 0 | false | true | 1.6229 | Follow-up: the previous reply did not fully solve the issue. Summarise this internal security policy into three bullet points for non-technical employees. | You are an internal knowledge assistant helping teammates navigate documentation, policies, and internal tools. Be precise, concise, and point to relevant information and next steps. | false | d33eca392020f2e6e14ec7fe3091a25b46d6bae1 | 212 | true |
INT000115 | S01055 | U00322 | 2025-04-11T03:47:38Z | train | web_app | coding_assistant | AU | APAC | free | individual | general_assistant | openai | gpt-4o-mini | 0.375 | 0.796 | 1,024 | 1,133 | 330 | 1,463 | 1,060 | 0.02928 | 0.757 | false | none | false | false | false | false | false | false | 0 | none | 2 | positive | false | 1 | decompose_task | abandoned | medium | Can you help me debug this? Follow-up: the previous reply did not fully solve the issue. Can you refactor this function to be more readable and efficient while keeping the same behaviour? | Let's turn this into a reliable workflow. Goal: handle async I/O pitfalls reliably, with guardrails for formatting and failure handling. Plan: - Start by writing a small function boundary for async I/O pitfalls and add... | 1,380.188679 | 2025-04-11 | 3 | 4 | false | false | 3.4333 | Follow-up: the previous reply did not fully solve the issue. Can you refactor this function to be more readable and efficient while keeping the same behaviour? | You are a senior software engineer helping users with code, debugging, and best practices. Provide clear explanations, structured steps, and high-quality code examples when relevant. | true | 3ddc6acb5986d68a027a4c760a5adfbe894b8deb | 221 | true |
INT000116 | S01236 | U00204 | 2025-03-31T16:55:34Z | train | internal_tool | internal_qa | GB | EMEA | pro | individual | support_strict | openai | gpt-4o-mini | 0.538 | 0.847 | 1,536 | 299 | 560 | 859 | 1,783 | 0.01716 | 0.853 | false | none | false | false | false | false | false | false | 0 | none | 2 | positive | false | 0 | none | auto_resolved | critical | Context for you: Summarise this internal security policy into three bullet points for non-technical employees. | Let's turn this into a reliable workflow. Goal: handle safety flags reliably, with guardrails for formatting and failure handling. Plan: - Start by writing a small function boundary for safety flags and add tests before... | 481.772294 | 2025-03-31 | 16 | 0 | false | true | 0.5339 | Summarise this internal security policy into three bullet points for non-technical employees. | You are an internal knowledge assistant helping teammates navigate documentation, policies, and internal tools. Be precise, concise, and point to relevant information and next steps. | true | 4e3bb3b7c597daf93de3f611632f4c56cefdec0e | 222 | true |
INT000117 | S01173 | U00262 | 2025-02-05T00:54:01Z | train | web_app | customer_support | IN | APAC | free | enterprise_team | creative_writer | local | custom-local-8b | 0.596 | 0.929 | 768 | 279 | 254 | 533 | 651 | 0.00212 | 0.608 | true | tool_error | false | false | false | false | true | false | 1 | tool_invocation_failed | -1 | negative | true | 1 | manual_post_edit | auto_resolved | low | Can you help me debug this? Follow-up: the previous reply did not fully solve the issue. Hi, I'm having an issue with my billing. The invoice total doesn't match what I expected. Can you check it? | Let's turn this into a reliable workflow. Let's stabilize it. Signal: tool_error while working on login issues. Quick diagnosis: Implement a fallback path when the tool is degraded. Recommended actions: - Implement retry with exponential... | 818.740399 | 2025-02-05 | 0 | 2 | false | false | 1.0984 | Follow-up: the previous reply did not fully solve the issue. Hi, I'm having an issue with my billing. The invoice total doesn't match what I expected. Can you check it? | You are a helpful customer support assistant for a SaaS product. Answer clearly, empathetically, and step-by-step while keeping the response concise and practical. | true | 7a490bfafd7436aea43a528e9fe4cb6db10b9a6b | 240 | true |
INT000118 | S00980 | U00421 | 2025-04-17T15:52:47Z | train | mobile_app | coding_assistant | US | AMER | free | individual | internal_policy_qa | google | gemini-1.5-pro | 0.101 | 0.873 | 1,536 | 1,058 | 387 | 1,445 | 1,434 | 0.4624 | 0.633 | true | tool_error | false | false | false | false | true | false | 1 | tool_invocation_failed | null | no_feedback | false | 2 | enable_tools | auto_resolved | low | Can you help me debug this? Follow-up: the previous reply did not fully solve the issue. I need help understanding why this asynchronous code is not awaiting correctly. Explain with a small code snippet. A short, clear answer would be ideal. | Let's make the behavior deterministic and observable. Signal: tool_error while working on async I/O pitfalls. Quick diagnosis: Separate tool telemetry from model telemetry so you can see which component failed.... | 1,007.670851 | 2025-04-17 | 15 | 3 | false | true | 2.7339 | Follow-up: the previous reply did not fully solve the issue. I need help understanding why this asynchronous code is not awaiting correctly. Explain with a small code snippet. | You are a senior software engineer helping users with code, debugging, and best practices. Provide clear explanations, structured steps, and high-quality code examples when relevant. | false | f84eff1764fee544ced395082209fcd5e43e6a0c | 213 | true |
INT000119 | S00740 | U00202 | 2025-02-03T16:50:12Z | train | web_app | brainstorming | AE | EMEA | pro | team | internal_policy_qa | openai | gpt-4.1 | 0.328 | 1 | 1,536 | 868 | 276 | 1,144 | 1,633 | 0.4576 | 0.927 | false | none | false | false | false | false | false | false | 0 | none | 1 | positive | false | 0 | none | auto_resolved | low | Quick question: Give me ten creative ideas for a YouTube series about learning data science from scratch. | We can make this robust with a few changes. Goal: handle developer productivity reliably, with guardrails for formatting and failure handling. Plan: - Add telemetry fields for latency/tokens/cost so you can debug... | 700.551133 | 2025-02-03 | 16 | 0 | false | true | 3.1449 | Give me ten creative ideas for a YouTube series about learning data science from scratch. | You are a creative assistant helping users explore ideas and options. Suggest multiple alternatives, structure them clearly, and stay practical. | true | 64f5ffde54eb84c78cab2d2147174c299cd15237 | 215 | true |
INT000120 | S01013 | U00042 | 2025-02-10T06:01:32Z | test | slack | coding_assistant | IN | APAC | free | individual | internal_policy_qa | meta | llama-3.1-70b | 0.186 | 0.925 | 1,536 | 342 | 203 | 545 | 1,002 | 0.109 | 0.767 | false | none | false | false | false | false | false | false | 2 | code_run|vector_search | 1 | positive | false | 0 | none | auto_resolved | high | Context for you: Can you refactor this function to be more readable and efficient while keeping the same behaviour? Can you walk me through a good solution? | Let's turn this into a reliable workflow. Goal: handle FastAPI endpoint design reliably, with guardrails for formatting and failure handling. Plan: - Prefer simple baselines first; then iterate based on measured failure slices. - Add telemetry... | 543.912176 | 2025-02-10 | 6 | 0 | false | false | 1.6847 | Can you refactor this function to be more readable and efficient while keeping the same behaviour? | You are a senior software engineer helping users with code, debugging, and best practices. Provide clear explanations, structured steps, and high-quality code examples when relevant. | true | f84eff1764fee544ced395082209fcd5e43e6a0c | 246 | true |
INT000121 | S00095 | U00069 | 2025-04-04T01:58:55Z | train | web_app | content_writing | IN | APAC | free | individual | internal_policy_qa | mistral | mistral-large | 0.643 | 0.896 | 768 | 743 | 138 | 881 | 782 | 0.10572 | 0.529 | true | toxicity | false | true | false | false | false | false | 0 | none | -1 | negative | true | 1 | switch_model_family | escalated_to_human | high | Context for you: Follow-up: the previous reply did not fully solve the issue. Draft a friendly email announcing a small price increase to our existing customers. | Here's the clean way to approach it. Let's stabilize it. Signal: toxicity while working on prompt injection. Quick diagnosis: Add a rewrite pathway: sanitize unsafe content into safe alternatives. Recommended actions: - Implement... | 1,126.598465 | 2025-04-04 | 1 | 4 | false | false | 5.3841 | Follow-up: the previous reply did not fully solve the issue. Draft a friendly email announcing a small price increase to our existing customers. | You are a writing assistant helping users craft clear, engaging, and well-structured content. Keep the tone natural, readable, and aligned with the user's intent. | true | db47a2ba00ac8f8120dcd2fd47d3648a8f594c66 | 232 | true |
INT000122 | S00611 | U00058 | 2025-04-16T02:31:50Z | train | mobile_app | internal_qa | IN | APAC | free | team | support_strict | meta | llama-3.1-8b | 0.723 | 0.917 | 1,536 | 788 | 345 | 1,133 | 660 | 0.068 | 0.63 | true | safety_block | false | false | true | false | false | false | 0 | none | -1 | negative | false | 0 | manual_post_edit | auto_resolved | low | I'm working on something and need help: Check this draft announcement for any confusing or inconsistent statements. | Below is a decision-focused breakdown. Let's stabilize it. Signal: safety_block while working on token spikes. Quick diagnosis: Route to a safe-help template and suggest benign resources or best practices. Recommended actions: - Add... | 1,716.666667 | 2025-04-16 | 2 | 2 | false | false | 2.2841 | Check this draft announcement for any confusing or inconsistent statements. | You are an internal knowledge assistant helping teammates navigate documentation, policies, and internal tools. Be precise, concise, and point to relevant information and next steps. | true | 4e3bb3b7c597daf93de3f611632f4c56cefdec0e | 235 | true |
INT000123 | S00196 | U00169 | 2025-02-03T16:15:44Z | train | web_app | content_writing | US | AMER | pro | individual | general_assistant | anthropic | claude-3.5-sonnet | 1.2 | 0.889 | 1,536 | 1,580 | 94 | 1,674 | 772 | 0.53568 | 1 | false | none | false | false | false | false | false | false | 0 | none | 2 | positive | false | 0 | none | auto_resolved | high | Draft a friendly email announcing a small price increase to our existing customers. Can you walk me through a good solution? | Here's the clean way to approach it. Goal: handle LLM reliability reliably, with guardrails for formatting and failure handling. Plan: - Use schema validation (e.g., Pydantic) to prevent bad inputs from... | 2,168.393782 | 2025-02-03 | 16 | 0 | false | true | 16.8085 | Draft a friendly email announcing a small price increase to our existing customers. | You are a writing assistant helping users craft clear, engaging, and well-structured content. Keep the tone natural, readable, and aligned with the user's intent. | true | a1c74d04f735d164fdc621e3ce3d5bfea1026243 | 205 | true |
INT000124 | S00648 | U00185 | 2025-04-24T13:06:09Z | train | api | data_analysis | US | AMER | pro | enterprise_team | creative_writer | openai | gpt-4.1 | 0.364 | 0.9 | 512 | 1,409 | 96 | 1,505 | 435 | 0.602 | 0.769 | false | none | false | false | false | false | false | false | 3 | code_run|email_send|vector_search | 1 | positive | false | 0 | none | auto_resolved | medium | I'm stuck on this and need some guidance: Can you suggest a few plots to explore churn risk in this dataset of subscriptions? | Here are the guardrails I'd put in place. Goal: handle tool-call success rate reliably, with guardrails for formatting and failure handling. Plan: - Use schema validation (e.g., Pydantic) to prevent bad inputs from propagating. - Start by writing a... | 3,459.770115 | 2025-04-24 | 13 | 3 | false | true | 14.6771 | Can you suggest a few plots to explore churn risk in this dataset of subscriptions? | You are a data analysis assistant helping users reason about data, metrics, and models. Explain trade-offs, outline steps, and keep the language accessible but precise. | true | 814335a175319858ca540d38bfcac020ec643c3f | 251 | true |
INT000125 | S00304 | U00128 | 2025-02-24T09:29:03Z | train | web_app | data_analysis | DE | EMEA | free | individual | creative_writer | google | gemini-1.5-pro | 0.141 | 0.928 | 512 | 786 | 30 | 816 | 589 | 0.26112 | 0.767 | false | none | false | false | false | false | false | false | 0 | none | null | no_feedback | false | 0 | none | auto_resolved | low | Can you help me debug this? I want a short summary of key metrics for this dataset: total users, active users, and average revenue per user. I'd like a practical, engineering-focused answer. | This is a common pattern; the fix is straightforward. Goal: handle safety flag rate reliably, with guardrails for formatting and failure handling. Plan: - Prefer simple baselines first; then iterate based on measured failure... | 1,385.398981 | 2025-02-24 | 9 | 0 | false | true | 26.2 | I want a short summary of key metrics for this dataset: total users, active users, and average revenue per user. | You are a data analysis assistant helping users reason about data, metrics, and models. Explain trade-offs, outline steps, and keep the language accessible but precise. | false | 814335a175319858ca540d38bfcac020ec643c3f | 227 | true |
INT000126 | S00823 | U00325 | 2025-02-19T20:37:52Z | train | mobile_app | coding_assistant | BR | AMER | enterprise | individual | code_assistant_secure | meta | llama-3.1-8b | 0.405 | 0.972 | 512 | 828 | 512 | 1,340 | 545 | 0.0804 | 0.842 | false | none | false | false | false | false | false | false | 0 | none | null | no_feedback | false | 0 | none | auto_resolved | high | Context for you: Can you refactor this function to be more readable and efficient while keeping the same behaviour? Can you walk me through a good solution? | Let's make the behavior deterministic and observable. Goal: handle JSON schema validation reliably, with guardrails for formatting and failure handling. Plan: - Prefer simple baselines first; then iterate based on measured failure slices. - Use schema... | 2,458.715596 | 2025-02-19 | 20 | 2 | false | false | 1.6172 | Can you refactor this function to be more readable and efficient while keeping the same behaviour? | You are a senior software engineer helping users with code, debugging, and best practices. Provide clear explanations, structured steps, and high-quality code examples when relevant. | false | 52eadb52b3319d2dab569ce8fb653a0bee56a90f | 254 | true |
INT000127 | S01461 | U00435 | 2025-03-06T22:00:48Z | val | slack | coding_assistant | BR | AMER | pro | team | internal_policy_qa | openai | gpt-4.1 | 0.291 | 0.819 | 768 | 592 | 225 | 817 | 649 | 0.3268 | 0.761 | false | none | false | false | false | false | false | false | 2 | email_send|vector_search | null | no_feedback | false | 0 | none | auto_resolved | medium | I'm stuck on this and need some guidance: Can you refactor this function to be more readable and efficient while keeping the same behaviour? Can you walk me through a good solution? | Here are the guardrails I'd put in place. Goal: handle async I/O pitfalls reliably, with guardrails for formatting and failure handling. Plan: - Prefer simple baselines first; then iterate based on measured failure slices. - Use... | 1,258.859784 | 2025-03-06 | 22 | 3 | false | false | 2.6311 | Can you refactor this function to be more readable and efficient while keeping the same behaviour? | You are a senior software engineer helping users with code, debugging, and best practices. Provide clear explanations, structured steps, and high-quality code examples when relevant. | false | f84eff1764fee544ced395082209fcd5e43e6a0c | 231 | true |
INT000128 | S00643 | U00363 | 2025-03-31T16:18:07Z | train | mobile_app | internal_qa | FR | EMEA | pro | individual | general_assistant | anthropic | claude-3.5-sonnet | 0.311 | 0.905 | 256 | 114 | 256 | 370 | 805 | 0.1184 | 0.511 | true | hallucination | true | false | false | false | false | false | 0 | none | 0 | neutral | false | 1 | regenerate_with_lower_temp | escalated_to_human | low | Can you help me debug this? Follow-up: the previous reply did not fully solve the issue. Turn this long internal announcement into a short FAQ-style document. Please explain step by step. | Here's the clean way to approach it. Signal: hallucination while working on safety flags. Quick diagnosis: Add grounding: retrieve evidence, require citations, and verify key facts. Recommended... | 459.627329 | 2025-03-31 | 16 | 0 | false | true | 0.4453 | Follow-up: the previous reply did not fully solve the issue. Turn this long internal announcement into a short FAQ-style document. | You are an internal knowledge assistant helping teammates navigate documentation, policies, and internal tools. Be precise, concise, and point to relevant information and next steps. | true | 7cd4f5f09ad56cc7a4cac3fea1f795d4e2ccc742 | 196 | true |
INT000129 | S01586 | U00385 | 2025-04-26T04:16:01Z | val | mobile_app | customer_support | EG | EMEA | free | team | creative_writer | anthropic | claude-3.5-sonnet | 0.375 | 1 | 1,024 | 129 | 490 | 619 | 1,678 | 0.19808 | 0.614 | true | formatting_error | false | false | false | true | false | false | 1 | email_send | null | no_feedback | false | 0 | manual_post_edit | auto_resolved | low | Quick question: A user says they were double charged this month. Help me draft a clear explanation and next steps. Please explain step by step. | Here's a clear, step-by-step answer. Signal: formatting_error while working on login issues. Quick diagnosis: Fail closed: if invalid, don't pass downstream; retry with the validator error. Recommended... | 368.891538 | 2025-04-26 | 4 | 5 | true | false | 0.2633 | A user says they were double charged this month. Help me draft a clear explanation and next steps. | You are a helpful customer support assistant for a SaaS product. Answer clearly, empathetically, and step-by-step while keeping the response concise and practical. | false | 7a490bfafd7436aea43a528e9fe4cb6db10b9a6b | 204 | true |
INT000130 | S01131 | U00289 | 2025-02-14T19:28:33Z | train | slack | coding_assistant | US | AMER | pro | individual | internal_policy_qa | meta | llama-3.1-8b | 0.258 | 0.912 | 1,536 | 640 | 933 | 1,573 | 495 | 0.0944 | 1 | false | none | false | false | false | false | false | false | 0 | none | 2 | positive | false | 0 | none | auto_resolved | high | Quick question: Write a small example that shows how to handle exceptions properly in Go. Can you walk me through a good solution? | This is a common pattern; the fix is straightforward. Goal: handle unit test mocking reliably, with guardrails for formatting and failure handling. Plan: - Use schema validation (e.g., Pydantic) to prevent bad inputs from propagating. - Prefer... | 3,177.777778 | 2025-02-14 | 19 | 4 | false | false | 0.686 | Write a small example that shows how to handle exceptions properly in Go. | You are a senior software engineer helping users with code, debugging, and best practices. Provide clear explanations, structured steps, and high-quality code examples when relevant. | true | f84eff1764fee544ced395082209fcd5e43e6a0c | 246 | true |
INT000131 | S00500 | U00292 | 2025-03-17T21:38:57Z | train | web_app | coding_assistant | US | AMER | enterprise | team | analytics_helper | local | custom-local-8b | 0.376 | 0.961 | 2,048 | 606 | 299 | 905 | 1,884 | 0.00364 | 0.925 | false | none | false | false | false | false | false | false | 1 | workflow_orchestrator | 0 | neutral | false | 0 | none | auto_resolved | low | I'm working on something and need help: I need help understanding why this asynchronous code is not awaiting correctly. Explain with a small code snippet. A short, clear answer would be ideal. | Let's make the behavior deterministic and observable. Goal: handle API retry logic reliably, with guardrails for formatting and failure handling. Plan: - Use schema validation (e.g., Pydantic) to prevent bad inputs from propagating. - Prefer simple... | 480.360934 | 2025-03-17 | 21 | 0 | false | false | 2.0268 | I need help understanding why this asynchronous code is not awaiting correctly. Explain with a small code snippet. | You are a senior software engineer helping users with code, debugging, and best practices. Provide clear explanations, structured steps, and high-quality code examples when relevant. | true | 54f94ac4f6df7ccc4808605ae4fb839b7a1fbfb2 | 251 | true |
INT000132 | S00541 | U00229 | 2025-02-14T08:31:13Z | train | api | coding_assistant | US | AMER | free | team | internal_policy_qa | mistral | mistral-large | 0.127 | 0.787 | 768 | 100 | 459 | 559 | 2,614 | 0.06708 | 0.808 | false | none | false | false | false | false | false | false | 3 | code_run|vector_search|workflow_orchestrator | 2 | positive | false | 0 | none | auto_resolved | medium | Quick question: I'm getting a JavaScript error: `ValueError`. Can you explain what it means and how to fix it? Please explain step by step. | I'll keep this concrete and measurable. Goal: handle API retry logic reliably, with guardrails for formatting and failure handling. Plan: - Add telemetry fields for latency/tokens/cost so you can debug regressions quickly. - Prefer simple... | 213.848508 | 2025-02-14 | 8 | 4 | false | false | 0.2179 | I'm getting a JavaScript error: `ValueError`. Can you explain what it means and how to fix it? | You are a senior software engineer helping users with code, debugging, and best practices. Provide clear explanations, structured steps, and high-quality code examples when relevant. | true | f84eff1764fee544ced395082209fcd5e43e6a0c | 241 | true |
INT000133 | S00552 | U00283 | 2025-03-31T09:54:06Z | val | mobile_app | customer_support | FR | EMEA | free | team | analytics_helper | openai | gpt-4o-mini | 0.584 | 0.885 | 768 | 262 | 152 | 414 | 351 | 0.00828 | 0.661 | false | none | false | false | false | false | false | false | 0 | none | 1 | positive | false | 0 | none | auto_resolved | low | Context for you: Hi, I'm having an issue with my billing. The invoice total doesn't match what I expected. Can you check it? I'd like a practical, engineering-focused answer. | Here's the clean way to approach it. Goal: handle data export request reliably, with guardrails for formatting and failure handling. Plan: - Add telemetry fields for latency/tokens/cost so you can debug regressions quickly. -... | 1,179.487179 | 2025-03-31 | 9 | 0 | false | true | 1.7237 | Hi, I'm having an issue with my billing. The invoice total doesn't match what I expected. Can you check it? | You are a helpful customer support assistant for a SaaS product. Answer clearly, empathetically, and step-by-step while keeping the response concise and practical. | true | 598eba0ae15195966a9835b68c542f0b34fb88e4 | 228 | true |
INT000134 | S00421 | U00411 | 2025-04-22T03:19:02Z | train | slack | data_analysis | IN | APAC | pro | team | internal_policy_qa | google | gemini-1.5-pro | 0.118 | 0.881 | 1,024 | 407 | 965 | 1,372 | 1,297 | 0.43904 | 1 | false | none | false | false | false | false | false | false | 2 | http_request|sql_query | 1 | positive | false | 0 | none | auto_resolved | critical | Context for you: Can you suggest a few plots to explore churn risk in this dataset of subscriptions? | Let's turn this into a reliable workflow. Goal: handle token usage reliably, with guardrails for formatting and failure handling. Plan: - Prefer simple baselines first; then iterate based on measured failure slices. - Add telemetry fields for... | 1,057.825752 | 2025-04-22 | 3 | 1 | false | false | 0.4218 | Can you suggest a few plots to explore churn risk in this dataset of subscriptions? | You are a data analysis assistant helping users reason about data, metrics, and models. Explain trade-offs, outline steps, and keep the language accessible but precise. | true | 62646e5a7026d4f68ab1867a730451326aa3b664 | 245 | true |
INT000135 | S01543 | U00364 | 2025-02-06T11:40:55Z | train | api | customer_support | US | AMER | enterprise | individual | code_assistant_secure | local | custom-local-8b | 0.621 | 0.828 | 1,536 | 1,253 | 76 | 1,329 | 1,131 | 0.00532 | 0.994 | false | none | false | false | false | false | false | false | 3 | email_send|knowledge_lookup|workflow_orchestrator | 2 | positive | false | 0 | none | auto_resolved | low | I'm stuck on this and need some guidance: Hi, I'm having an issue with my billing. The invoice total doesn't match what I expected. Can you check it? Please explain step by step. | Here are the guardrails I'd put in place. Goal: handle login issues reliably, with guardrails for formatting and failure handling. Plan: - Use schema validation (e.g., Pydantic) to prevent bad inputs from propagating. - Start by... | 1,175.066313 | 2025-02-06 | 11 | 3 | false | true | 16.4868 | Hi, I'm having an issue with my billing. The invoice total doesn't match what I expected. Can you check it? | You are a helpful customer support assistant for a SaaS product. Answer clearly, empathetically, and step-by-step while keeping the response concise and practical. | true | 1a8e13c1d6c8047debb9a230bf3ae08524ccb884 | 231 | true |
INT000136 | S00229 | U00180 | 2025-03-18T00:55:04Z | train | mobile_app | customer_support | IN | APAC | pro | individual | analytics_helper | google | gemini-1.5-pro | 0.31 | 0.916 | 768 | 1,379 | 395 | 1,774 | 848 | 0.56768 | 0.303 | true | hallucination | true | false | false | false | false | false | 0 | none | null | no_feedback | true | 0 | decompose_task | escalated_to_human | low | Our customer is getting an error when trying to log in to their account. Can you help me debug this? Please explain step by step. | Here's a clear, step-by-step answer. Signal: hallucination while working on data export request. Quick diagnosis: Add grounding: retrieve evidence, require citations, and verify key facts.... | 2,091.981132 | 2025-03-18 | 0 | 1 | false | false | 3.4911 | Our customer is getting an error when trying to log in to their account. Can you help me debug this? | You are a helpful customer support assistant for a SaaS product. Answer clearly, empathetically, and step-by-step while keeping the response concise and practical. | false | 598eba0ae15195966a9835b68c542f0b34fb88e4 | 191 | true |
INT000137 | S00023 | U00045 | 2025-03-05T16:24:08Z | train | mobile_app | data_analysis | BR | AMER | enterprise | individual | code_assistant_secure | local | custom-local-8b | 0.241 | 0.876 | 1,536 | 1,262 | 115 | 1,377 | 951 | 0.00552 | 0.81 | false | none | false | false | false | false | false | false | 2 | code_run|knowledge_lookup | 2 | positive | false | 0 | none | auto_resolved | critical | I'm stuck on this and need some guidance: Help me compute the conversion rate per marketing channel and highlight the top 3 performers. Please explain step by step. | Let's make the behavior deterministic and observable. Goal: handle tool-call success rate reliably, with guardrails for formatting and failure handling. Plan: - Use schema validation (e.g., Pydantic) to prevent bad inputs from propagating. - Start by... | 1,447.949527 | 2025-03-05 | 16 | 2 | false | true | 10.9739 | Help me compute the conversion rate per marketing channel and highlight the top 3 performers. | You are a data analysis assistant helping users reason about data, metrics, and models. Explain trade-offs, outline steps, and keep the language accessible but precise. | true | 4e275efe35fdc513983e7d3f3ad32b50cc391969 | 253 | true |
INT000138 | S00484 | U00177 | 2025-02-23T15:32:30Z | train | api | data_analysis | US | AMER | free | individual | creative_writer | local | custom-local-8b | 0.218 | 0.979 | 256 | 828 | 256 | 1,084 | 985 | 0.00432 | 0.979 | false | none | false | false | false | false | false | false | 3 | email_send|http_request|vector_search | 2 | positive | false | 0 | none | auto_resolved | medium | I'm working on something and need help: I have a CSV with user sessions. Show me a quick breakdown of sessions per country and device type. I'd like a practical, engineering-focused answer. | Here's the clean way to approach it. Goal: handle latency p95 reliably, with guardrails for formatting and failure handling. Plan: - Prefer simple baselines first; then iterate based on measured failure slices. - Use schema... | 1,100.507614 | 2025-02-23 | 15 | 6 | true | true | 3.2344 | I have a CSV with user sessions. Show me a quick breakdown of sessions per country and device type. | You are a data analysis assistant helping users reason about data, metrics, and models. Explain trade-offs, outline steps, and keep the language accessible but precise. | true | 814335a175319858ca540d38bfcac020ec643c3f | 226 | true |
INT000139 | S01458 | U00386 | 2025-04-19T00:16:48Z | test | web_app | data_analysis | IN | APAC | free | individual | code_assistant_secure | local | custom-local-8b | 0.239 | 0.939 | 1,024 | 23 | 699 | 722 | 644 | 0.00288 | 0.804 | false | none | false | false | false | false | false | false | 0 | none | null | no_feedback | false | 0 | none | auto_resolved | medium | Context for you: I have a CSV with user sessions. Show me a quick breakdown of sessions per country and device type. A short, clear answer would be ideal. | Here's a clear, step-by-step answer. Goal: handle safety flag rate reliably, with guardrails for formatting and failure handling. Plan: - Prefer simple baselines first; then iterate based on measured failure slices. - Start by writing a... | 1,121.118012 | 2025-04-19 | 0 | 5 | true | false | 0.0329 | I have a CSV with user sessions. Show me a quick breakdown of sessions per country and device type. | You are a data analysis assistant helping users reason about data, metrics, and models. Explain trade-offs, outline steps, and keep the language accessible but precise. | false | 4e275efe35fdc513983e7d3f3ad32b50cc391969 | 239 | true |
INT000140 | S01372 | U00354 | 2025-02-25T09:50:49Z | train | mobile_app | customer_support | IN | APAC | enterprise | individual | support_strict | openai | gpt-4.1 | 0.314 | 0.72 | 1,024 | 907 | 909 | 1,816 | 1,450 | 0.7264 | 0.739 | false | none | false | false | false | false | false | false | 0 | none | 0 | neutral | false | 0 | none | escalated_to_human | low | I'm stuck on this and need some guidance: Hi, I'm having an issue with my billing. The invoice total doesn't match what I expected. Can you check it? A short, clear answer would be ideal. | Here's a clear, step-by-step answer. Goal: handle login issues reliably, with guardrails for formatting and failure handling. Plan: - Use schema validation (e.g., Pydantic) to prevent bad inputs from propagating. - Prefer simple baselines... | 1,252.413793 | 2025-02-25 | 9 | 1 | false | true | 0.9978 | Hi, I'm having an issue with my billing. The invoice total doesn't match what I expected. Can you check it? | You are a helpful customer support assistant for a SaaS product. Answer clearly, empathetically, and step-by-step while keeping the response concise and practical. | true | d029d0919f8e306d09ca7001d500c1e6f1c81007 | 241 | true |
INT000141 | S01281 | U00343 | 2025-04-20T19:43:32Z | test | api | coding_assistant | US | AMER | enterprise | individual | internal_policy_qa | google | gemini-1.5-pro | 0.386 | 0.952 | 512 | 1,676 | 384 | 2,060 | 2,001 | 0.6592 | 0.804 | false | none | false | false | false | false | false | false | 0 | none | null | no_feedback | false | 1 | add_constraints_to_prompt | auto_resolved | low | Follow-up: the previous reply did not fully solve the issue. Can you refactor this function to be more readable and efficient while keeping the same behaviour? A short, clear answer would be ideal. | Here's what I'd implement in production. Goal: handle JSON schema validation reliably, with guardrails for formatting and failure handling. Plan: - Add telemetry fields for latency/tokens/cost so you can debug... | 1,029.485257 | 2025-04-20 | 19 | 6 | true | false | 4.3646 | Follow-up: the previous reply did not fully solve the issue. Can you refactor this function to be more readable and efficient while keeping the same behaviour? | You are a senior software engineer helping users with code, debugging, and best practices. Provide clear explanations, structured steps, and high-quality code examples when relevant. | false | f84eff1764fee544ced395082209fcd5e43e6a0c | 212 | true |
INT000142 | S00110 | U00228 | 2025-02-04T14:13:56Z | train | mobile_app | customer_support | US | AMER | free | individual | internal_policy_qa | openai | gpt-4o-mini | 0.319 | 0.916 | 2,048 | 208 | 46 | 254 | 1,406 | 0.00508 | 0.918 | false | none | false | false | false | false | false | false | 0 | none | 1 | positive | false | 0 | none | escalated_to_human | medium | Quick question: Hi, I'm having an issue with my billing. The invoice total doesn't match what I expected. Can you check it? Please explain step by step. | Here's what I'd implement in production. Goal: handle data export request reliably, with guardrails for formatting and failure handling. Plan: - Add telemetry fields for latency/tokens/cost so you can debug regressions... | 180.654339 | 2025-02-04 | 14 | 1 | false | true | 4.5217 | Hi, I'm having an issue with my billing. The invoice total doesn't match what I expected. Can you check it? | You are a helpful customer support assistant for a SaaS product. Answer clearly, empathetically, and step-by-step while keeping the response concise and practical. | true | 03bf57d7adc83026d525c8192379aa203b71e1eb | 221 | true |
INT000143 | S01211 | U00405 | 2025-04-05T22:45:54Z | train | api | content_writing | CA | AMER | free | team | analytics_helper | mistral | mistral-large | 0.87 | 0.826 | 768 | 268 | 768 | 1,036 | 1,681 | 0.12432 | 0.796 | false | none | false | false | false | false | false | false | 0 | none | 2 | positive | false | 0 | none | auto_resolved | low | Quick question: Write a concise product description for a new wireless mouse aimed at software developers. I'd like a practical, engineering-focused answer. | Let's break this down into actionable steps. Goal: handle LLM reliability reliably, with guardrails for formatting and failure handling. Plan: - Add telemetry fields for latency/tokens/cost so you can debug regressions quickly. - Start... | 616.299822 | 2025-04-05 | 22 | 5 | true | false | 0.349 | Write a concise product description for a new wireless mouse aimed at software developers. | You are a writing assistant helping users craft clear, engaging, and well-structured content. Keep the tone natural, readable, and aligned with the user's intent. | true | 9716fa3b2c6fe39ac0f75ddaad781dfa757a0de9 | 238 | true |
INT000144 | S01466 | U00020 | 2025-02-05T09:50:50Z | train | web_app | data_analysis | EG | EMEA | free | individual | creative_writer | anthropic | claude-3.5-sonnet | 0.329 | 0.812 | 512 | 603 | 12 | 615 | 1,058 | 0.1968 | 0.766 | false | none | false | false | false | false | false | false | 1 | code_run | 1 | positive | false | 0 | none | auto_resolved | medium | I'm working on something and need help: Can you suggest a few plots to explore churn risk in this dataset of subscriptions? | This is a common pattern; the fix is straightforward. Goal: handle token usage reliably, with guardrails for formatting and failure handling. Plan: - Use schema validation (e.g., Pydantic) to prevent bad inputs from propagating. - Add telemetry... | 581.285444 | 2025-02-05 | 9 | 2 | false | true | 50.25 | Can you suggest a few plots to explore churn risk in this dataset of subscriptions? | You are a data analysis assistant helping users reason about data, metrics, and models. Explain trade-offs, outline steps, and keep the language accessible but precise. | true | 814335a175319858ca540d38bfcac020ec643c3f | 247 | true |
INT000145 | S00792 | U00078 | 2025-04-26T23:17:00Z | test | web_app | data_analysis | US | AMER | pro | team | code_assistant_secure | openai | gpt-4o-mini | 0.143 | 0.858 | 768 | 928 | 748 | 1,676 | 455 | 0.03352 | 1 | false | none | false | false | false | false | false | false | 0 | none | null | no_feedback | false | 0 | none | auto_resolved | medium | I'm stuck on this and need some guidance: I have a CSV with user sessions. Show me a quick breakdown of sessions per country and device type. Please explain step by step. | Here's the clean way to approach it. Goal: handle tool-call success rate reliably, with guardrails for formatting and failure handling. Plan: - Add telemetry fields for latency/tokens/cost so you can debug regressions quickly.... | 3,683.516484 | 2025-04-26 | 23 | 5 | true | false | 1.2406 | I have a CSV with user sessions. Show me a quick breakdown of sessions per country and device type. | You are a data analysis assistant helping users reason about data, metrics, and models. Explain trade-offs, outline steps, and keep the language accessible but precise. | false | 4e275efe35fdc513983e7d3f3ad32b50cc391969 | 229 | true |
INT000146 | S00981 | U00258 | 2025-04-10T16:33:54Z | train | api | customer_support | GB | EMEA | free | team | support_strict | anthropic | claude-3.5-sonnet | 0.576 | 0.917 | 512 | 443 | 317 | 760 | 2,071 | 0.2432 | 0.868 | false | none | false | false | false | false | false | false | 0 | none | null | no_feedback | false | 0 | none | auto_resolved | medium | Can you help me debug this? Our customer is getting an error when trying to log in to their account. Can you help me debug this? Please explain step by step. | Below is a decision-focused breakdown. Goal: handle account verification reliably, with guardrails for formatting and failure handling. Plan: - Add telemetry fields for latency/tokens/cost so you can debug... | 366.972477 | 2025-04-10 | 16 | 3 | false | true | 1.3975 | Our customer is getting an error when trying to log in to their account. Can you help me debug this? | You are a helpful customer support assistant for a SaaS product. Answer clearly, empathetically, and step-by-step while keeping the response concise and practical. | false | d029d0919f8e306d09ca7001d500c1e6f1c81007 | 208 | true |
INT000147 | S01252 | U00157 | 2025-03-23T07:54:36Z | train | mobile_app | coding_assistant | EG | EMEA | free | individual | analytics_helper | openai | gpt-4o-mini | 0.196 | 0.909 | 1,024 | 708 | 122 | 830 | 1,030 | 0.0166 | 0.47 | true | hallucination | true | false | false | false | false | false | 0 | none | -2 | negative | true | 1 | enable_tools | escalated_to_human | low | Can you help me debug this? Follow-up: the previous reply did not fully solve the issue. Write a small example that shows how to handle exceptions properly in Go. A short, clear answer would be ideal. | Here are the guardrails I'd put in place. Signal: hallucination while working on unit test mocking. Quick diagnosis: Add grounding: retrieve evidence, require citations, and verify key facts. Recommended... | 805.825243 | 2025-03-23 | 7 | 6 | true | false | 5.8033 | Follow-up: the previous reply did not fully solve the issue. Write a small example that shows how to handle exceptions properly in Go. | You are a senior software engineer helping users with code, debugging, and best practices. Provide clear explanations, structured steps, and high-quality code examples when relevant. | true | 54f94ac4f6df7ccc4808605ae4fb839b7a1fbfb2 | 206 | true |
INT000148 | S01012 | U00449 | 2025-03-25T21:46:17Z | train | slack | customer_support | BR | AMER | free | enterprise_team | creative_writer | openai | gpt-4o-mini | 0.653 | 0.934 | 1,024 | 829 | 452 | 1,281 | 1,723 | 0.02564 | 0.832 | false | none | false | false | false | false | false | false | 3 | code_run|knowledge_lookup|sql_query | 2 | positive | false | 1 | add_constraints_to_prompt | auto_resolved | low | Can you help me debug this? Follow-up: the previous reply did not fully solve the issue. Our customer is getting an error when trying to log in to their account. Can you help me debug this? | This is a common pattern; the fix is straightforward. Goal: handle unexpected charge reliably, with guardrails for formatting and failure handling. Plan: - Use schema validation (e.g., Pydantic) to prevent bad inputs from... | 743.470691 | 2025-03-25 | 21 | 1 | false | false | 1.8341 | Follow-up: the previous reply did not fully solve the issue. Our customer is getting an error when trying to log in to their account. Can you help me debug this? | You are a helpful customer support assistant for a SaaS product. Answer clearly, empathetically, and step-by-step while keeping the response concise and practical. | true | 7a490bfafd7436aea43a528e9fe4cb6db10b9a6b | 224 | true |
INT000149 | S01238 | U00037 | 2025-04-04T16:55:31Z | train | web_app | customer_support | FR | EMEA | enterprise | enterprise_team | internal_policy_qa | meta | llama-3.1-70b | 0.797 | 0.937 | 256 | 778 | 256 | 1,034 | 1,228 | 0.2068 | 0.903 | false | none | false | false | false | false | false | false | 2 | knowledge_lookup|sql_query | 2 | positive | false | 0 | none | auto_resolved | medium | I'm working on something and need help: Our customer is getting an error when trying to log in to their account. Can you help me debug this? Can you walk me through a good solution? | Here's a clear, step-by-step answer. Goal: handle login issues reliably, with guardrails for formatting and failure handling. Plan: - Add telemetry fields for latency/tokens/cost so you can debug regressions quickly. - Prefer simple... | 842.019544 | 2025-04-04 | 16 | 4 | false | true | 3.0391 | Our customer is getting an error when trying to log in to their account. Can you help me debug this? | You are a helpful customer support assistant for a SaaS product. Answer clearly, empathetically, and step-by-step while keeping the response concise and practical. | true | 03bf57d7adc83026d525c8192379aa203b71e1eb | 235 | true |
INT000150 | S01302 | U00152 | 2025-04-15T09:16:09Z | train | mobile_app | customer_support | GB | EMEA | free | team | general_assistant | openai | gpt-4o-mini | 0.88 | 0.822 | 512 | 286 | 418 | 704 | 1,283 | 0.01408 | 0.824 | false | none | false | false | false | false | false | false | 3 | code_run|knowledge_lookup|sql_query | 1 | positive | false | 0 | none | abandoned | high | A user says they were double charged this month. Help me draft a clear explanation and next steps. I'd like a practical, engineering-focused answer. | We can make this robust with a few changes. Goal: handle data export request reliably, with guardrails for formatting and failure handling. Plan: - Use schema validation (e.g., Pydantic) to prevent bad inputs from propagating. - Start by writing a... | 548.713952 | 2025-04-15 | 9 | 1 | false | true | 0.6842 | A user says they were double charged this month. Help me draft a clear explanation and next steps. | You are a helpful customer support assistant for a SaaS product. Answer clearly, empathetically, and step-by-step while keeping the response concise and practical. | true | a52fb80193c7c0769c0836852624f4b025a04d2a | 250 | true |
INT000151 | S01171 | U00075 | 2025-03-29T06:39:07Z | test | api | data_analysis | IN | APAC | pro | individual | support_strict | local | custom-local-8b | 0.202 | 0.811 | 1,024 | 123 | 232 | 355 | 545 | 0.00144 | 0.593 | true | toxicity | false | true | false | false | false | false | 0 | none | -1 | negative | true | 1 | regenerate_with_lower_temp | auto_resolved | medium | Can you help me debug this? Follow-up: the previous reply did not fully solve the issue. Help me compute the conversion rate per marketing channel and highlight the top 3 performers. Please explain step by step. | I'll keep this concrete and measurable. Signal: toxicity while working on cost per request. Quick diagnosis: Apply safety filtering and tighten refusal rules; log safety flags for monitoring. Recommended... | 651.376147 | 2025-03-29 | 6 | 5 | true | false | 0.5302 | Follow-up: the previous reply did not fully solve the issue. Help me compute the conversion rate per marketing channel and highlight the top 3 performers. | You are a data analysis assistant helping users reason about data, metrics, and models. Explain trade-offs, outline steps, and keep the language accessible but precise. | true | a15ac130f4f428f0efac738d5c68655674fccc6e | 206 | true |
INT000152 | S00058 | U00014 | 2025-03-29T01:57:18Z | train | web_app | coding_assistant | IN | APAC | pro | individual | code_assistant_secure | anthropic | claude-3.5-sonnet | 0.218 | 0.894 | 768 | 869 | 35 | 904 | 452 | 0.28928 | 0.891 | false | none | false | false | false | false | false | false | 0 | none | 1 | positive | false | 0 | none | auto_resolved | medium | Quick question: I need help understanding why this asynchronous code is not awaiting correctly. Explain with a small code snippet. | Here's what I'd implement in production. Goal: handle API retry logic reliably, with guardrails for formatting and failure handling. Plan: - Use schema validation (e.g., Pydantic) to prevent bad inputs from... | 2,000 | 2025-03-29 | 1 | 5 | true | false | 24.8286 | I need help understanding why this asynchronous code is not awaiting correctly. Explain with a small code snippet. | You are a senior software engineer helping users with code, debugging, and best practices. Provide clear explanations, structured steps, and high-quality code examples when relevant. | true | 52eadb52b3319d2dab569ce8fb653a0bee56a90f | 209 | true |
INT000153 | S00984 | U00163 | 2025-04-09T14:53:50Z | train | web_app | content_writing | FR | EMEA | pro | team | general_assistant | meta | llama-3.1-8b | 0.615 | 0.983 | 768 | 378 | 368 | 746 | 1,061 | 0.04476 | 0.643 | false | none | false | false | false | false | false | false | 0 | none | null | no_feedback | false | 0 | none | auto_resolved | high | Context for you: Rewrite this paragraph in a clearer and more professional tone. Can you walk me through a good solution? | Here's a clear, step-by-step answer. Goal: handle prompt injection reliably, with guardrails for formatting and failure handling. Plan: - Prefer simple baselines first; then iterate based on measured failure slices. - Add... | 703.110273 | 2025-04-09 | 14 | 2 | false | true | 1.0272 | Rewrite this paragraph in a clearer and more professional tone. | You are a writing assistant helping users craft clear, engaging, and well-structured content. Keep the tone natural, readable, and aligned with the user's intent. | false | a1c74d04f735d164fdc621e3ce3d5bfea1026243 | 224 | true |
INT000154 | S00488 | U00289 | 2025-04-30T16:10:02Z | train | internal_tool | content_writing | US | AMER | pro | individual | analytics_helper | local | custom-local-8b | 0.762 | 0.919 | 512 | 1,120 | 512 | 1,632 | 1,699 | 0.00652 | 0.931 | false | none | false | false | false | false | false | false | 0 | none | null | no_feedback | false | 0 | none | auto_resolved | low | Quick question: Draft a friendly email announcing a small price increase to our existing customers. | I'll keep this concrete and measurable. Goal: handle tool calling reliably, with guardrails for formatting and failure handling. Plan: - Prefer simple baselines first; then iterate based on measured failure slices. - Start by writing a small... | 960.565038 | 2025-04-30 | 16 | 2 | false | true | 2.1875 | Draft a friendly email announcing a small price increase to our existing customers. | You are a writing assistant helping users craft clear, engaging, and well-structured content. Keep the tone natural, readable, and aligned with the user's intent. | false | 9716fa3b2c6fe39ac0f75ddaad781dfa757a0de9 | 244 | true |
INT000155 | S00166 | U00267 | 2025-04-13T21:55:23Z | test | api | content_writing | IN | APAC | pro | individual | code_assistant_secure | openai | gpt-4o-mini | 0.644 | 0.825 | 512 | 490 | 512 | 1,002 | 2,090 | 0.02004 | 0.564 | true | hallucination | true | false | false | false | false | false | 0 | none | -1 | negative | false | 0 | manual_post_edit | auto_resolved | low | Context for you: Write a concise product description for a new wireless mouse aimed at software developers. Please explain step by step. | Below is a decision-focused breakdown. Signal: hallucination while working on A/B testing. Quick diagnosis: Refuse when confidence is low; avoid confident guesses. Recommended actions: - Implement retry... | 479.425837 | 2025-04-13 | 21 | 6 | true | false | 0.957 | Write a concise product description for a new wireless mouse aimed at software developers. | You are a writing assistant helping users craft clear, engaging, and well-structured content. Keep the tone natural, readable, and aligned with the user's intent. | true | 20f1dad1d39f7b8df6faaa166fe78024f8d03fd6 | 205 | true |
INT000156 | S00473 | U00305 | 2025-04-30T14:42:09Z | train | web_app | content_writing | GB | EMEA | pro | team | creative_writer | google | gemini-1.5-pro | 0.815 | 0.954 | 768 | 1,045 | 104 | 1,149 | 1,563 | 0.36768 | 0.81 | false | none | false | false | false | false | false | false | 0 | none | 2 | positive | false | 0 | none | auto_resolved | low | Quick question: Write a concise product description for a new wireless mouse aimed at software developers. | Here's a clear, step-by-step answer. Goal: handle prompt injection reliably, with guardrails for formatting and failure handling. Plan: - Use schema validation (e.g., Pydantic) to prevent bad inputs from... | 735.12476 | 2025-04-30 | 14 | 2 | false | true | 10.0481 | Write a concise product description for a new wireless mouse aimed at software developers. | You are a writing assistant helping users craft clear, engaging, and well-structured content. Keep the tone natural, readable, and aligned with the user's intent. | true | 469eb15f8b08d9b4130e14a6c3888f14e335a854 | 206 | true |
INT000157 | S01600 | U00071 | 2025-02-27T06:50:04Z | train | web_app | coding_assistant | GB | EMEA | pro | individual | analytics_helper | google | gemini-1.5-pro | 0.25 | 0.88 | 1,024 | 1,399 | 563 | 1,962 | 903 | 0.62784 | 0.605 | true | toxicity | false | true | false | false | false | false | 0 | none | null | no_feedback | false | 0 | regenerate_with_lower_temp | escalated_to_human | low | Context for you: I need help understanding why this asynchronous code is not awaiting correctly. Explain with a small code snippet. A short, clear answer would be ideal. | Let's turn this into a reliable workflow. Signal: toxicity while working on rate limiting. Quick diagnosis: Apply safety filtering and tighten refusal rules; log safety flags for monitoring.... | 2,172.757475 | 2025-02-27 | 6 | 3 | false | false | 2.4849 | I need help understanding why this asynchronous code is not awaiting correctly. Explain with a small code snippet. | You are a senior software engineer helping users with code, debugging, and best practices. Provide clear explanations, structured steps, and high-quality code examples when relevant. | false | 54f94ac4f6df7ccc4808605ae4fb839b7a1fbfb2 | 193 | true |
INT000158 | S00128 | U00146 | 2025-02-03T20:12:16Z | test | api | brainstorming | US | AMER | free | enterprise_team | analytics_helper | anthropic | claude-3.5-sonnet | 0.44 | 0.97 | 768 | 695 | 703 | 1,398 | 913 | 0.44736 | 0.972 | false | none | false | false | false | false | false | false | 0 | none | 2 | positive | false | 0 | none | auto_resolved | medium | I'm working on something and need help: Suggest a few content ideas for our next newsletter aimed at early-stage startups. A short, clear answer would be ideal. | Here's what I'd implement in production. Goal: handle customer support automation reliably, with guardrails for formatting and failure handling. Plan: - Use schema validation (e.g., Pydantic) to prevent bad inputs from propagating.... | 1,531.215772 | 2025-02-03 | 20 | 0 | false | false | 0.9886 | Suggest a few content ideas for our next newsletter aimed at early-stage startups. | You are a creative assistant helping users explore ideas and options. Suggest multiple alternatives, structure them clearly, and stay practical. | true | 2df52d847c72d1f6b37061123acb53282123086b | 234 | true |
INT000159 | S00842 | U00016 | 2025-02-20T09:59:56Z | train | web_app | customer_support | GB | EMEA | free | individual | internal_policy_qa | mistral | mistral-large | 0.63 | 1 | 2,048 | 113 | 146 | 259 | 1,658 | 0.03108 | 0.919 | false | none | false | false | false | false | false | false | 3 | code_run|sql_query|workflow_orchestrator | null | no_feedback | false | 0 | none | auto_resolved | medium | Quick question: A user says they were double charged this month. Help me draft a clear explanation and next steps. Please explain step by step. | Here's a clear, step-by-step answer. Goal: handle slow responses reliably, with guardrails for formatting and failure handling. Plan: - Prefer simple baselines first; then iterate based on measured failure slices. - Add telemetry fields for... | 156.212304 | 2025-02-20 | 9 | 3 | false | true | 0.774 | A user says they were double charged this month. Help me draft a clear explanation and next steps. | You are a helpful customer support assistant for a SaaS product. Answer clearly, empathetically, and step-by-step while keeping the response concise and practical. | false | 03bf57d7adc83026d525c8192379aa203b71e1eb | 243 | true |
INT000160 | S01276 | U00027 | 2025-04-20T16:05:06Z | test | api | coding_assistant | GB | EMEA | free | team | support_strict | meta | llama-3.1-8b | 0.308 | 0.883 | 768 | 233 | 416 | 649 | 1,803 | 0.03896 | 0.943 | false | none | false | false | false | false | false | false | 0 | none | null | no_feedback | false | 0 | none | auto_resolved | low | Can you help me debug this? I need help understanding why this asynchronous code is not awaiting correctly. Explain with a small code snippet. | We can make this robust with a few changes. Goal: handle unit test mocking reliably, with guardrails for formatting and failure handling. Plan: - Use schema validation (e.g., Pydantic) to prevent bad inputs from propagating. - Start by writing a... | 359.95563 | 2025-04-20 | 16 | 6 | true | true | 0.5601 | I need help understanding why this asynchronous code is not awaiting correctly. Explain with a small code snippet. | You are a senior software engineer helping users with code, debugging, and best practices. Provide clear explanations, structured steps, and high-quality code examples when relevant. | false | da34aac46ee8d6d44012043be92e33884c6b4920 | 248 | true |
INT000161 | S01073 | U00085 | 2025-02-15T13:25:55Z | test | mobile_app | brainstorming | US | AMER | free | enterprise_team | general_assistant | mistral | mistral-large | 0.381 | 0.859 | 1,024 | 1,097 | 825 | 1,922 | 1,259 | 0.23064 | 0.477 | true | tool_error | false | false | false | false | true | false | 1 | tool_invocation_failed | -2 | negative | true | 0 | decompose_task | abandoned | low | Can you help me debug this? I need naming ideas for an internal tool that tracks experiments and model runs. | Here are the guardrails I'd put in place. Signal: tool_error while working on developer productivity. Quick diagnosis: Validate tool arguments before calling the tool; reject invalid calls early.... | 1,526.608419 | 2025-02-15 | 13 | 5 | true | true | 1.3297 | I need naming ideas for an internal tool that tracks experiments and model runs. | You are a creative assistant helping users explore ideas and options. Suggest multiple alternatives, structure them clearly, and stay practical. | true | 62cac6cf36e4ae3f199823b7f2ca46aa5e4a0aa7 | 198 | true |
INT000162 | S01231 | U00207 | 2025-03-03T13:02:52Z | train | api | data_analysis | EG | EMEA | free | individual | creative_writer | google | gemini-1.5-pro | 0.137 | 0.931 | 768 | 718 | 567 | 1,285 | 934 | 0.4112 | 0.719 | false | none | false | false | false | false | false | false | 0 | none | 2 | positive | false | 0 | none | auto_resolved | high | Quick question: I have a CSV with user sessions. Show me a quick breakdown of sessions per country and device type. I'd like a practical, engineering-focused answer. | We can make this robust with a few changes. Goal: handle cost per request reliably, with guardrails for formatting and failure handling. Plan: - Add telemetry fields for latency/tokens/cost so you can debug regressions quickly. - Prefer simple... | 1,375.802998 | 2025-03-03 | 13 | 0 | false | true | 1.2663 | I have a CSV with user sessions. Show me a quick breakdown of sessions per country and device type. | You are a data analysis assistant helping users reason about data, metrics, and models. Explain trade-offs, outline steps, and keep the language accessible but precise. | true | 814335a175319858ca540d38bfcac020ec643c3f | 246 | true |
INT000163 | S00061 | U00054 | 2025-03-20T08:44:21Z | train | internal_tool | coding_assistant | GB | EMEA | free | team | code_assistant_secure | openai | gpt-4.1 | 0.244 | 0.867 | 1,536 | 693 | 356 | 1,049 | 864 | 0.4196 | 0.543 | true | safety_block | false | false | true | false | false | false | 0 | none | null | no_feedback | false | 3 | add_constraints_to_prompt | auto_resolved | high | Context for you: Follow-up: the previous reply did not fully solve the issue. I'm getting a Python error: `IndexError`. Can you explain what it means and how to fix it? Can you walk me through a good solution? | We can make this robust with a few changes. Signal: safety_block while working on caching strategy. Quick diagnosis: Refuse safely and offer a legitimate alternative or a high-level explanation.... | 1,214.12037 | 2025-03-20 | 8 | 3 | false | false | 1.9466 | Follow-up: the previous reply did not fully solve the issue. I'm getting a Python error: `IndexError`. Can you explain what it means and how to fix it? | You are a senior software engineer helping users with code, debugging, and best practices. Provide clear explanations, structured steps, and high-quality code examples when relevant. | false | 52eadb52b3319d2dab569ce8fb653a0bee56a90f | 197 | true |
INT000164 | S00867 | U00001 | 2025-02-24T01:34:16Z | test | api | internal_qa | IN | APAC | free | individual | internal_policy_qa | openai | gpt-4.1 | 0.591 | 0.921 | 2,048 | 80 | 493 | 573 | 834 | 0.2292 | 0.602 | true | tool_error | false | false | false | false | true | false | 1 | tool_invocation_failed | null | no_feedback | true | 3 | add_constraints_to_prompt | auto_resolved | low | Can you help me debug this? Follow-up: the previous reply did not fully solve the issue. Turn this long internal announcement into a short FAQ-style document. | Let's make the behavior deterministic and observable. Let's stabilize it. Signal: tool_error while working on errors. Quick diagnosis: Validate tool arguments before calling the tool; reject invalid calls early. Recommended actions: - Use caching... | 687.05036 | 2025-02-24 | 1 | 0 | false | false | 0.1623 | Follow-up: the previous reply did not fully solve the issue. Turn this long internal announcement into a short FAQ-style document. | You are an internal knowledge assistant helping teammates navigate documentation, policies, and internal tools. Be precise, concise, and point to relevant information and next steps. | false | 8ad879fd7895b38236ec6ff25a956251dc06e51b | 249 | true |
INT000165 | S00396 | U00424 | 2025-02-12T07:48:11Z | train | api | data_analysis | FR | EMEA | free | team | internal_policy_qa | openai | gpt-4o-mini | 0.196 | 0.841 | 1,536 | 739 | 763 | 1,502 | 966 | 0.03004 | 0.662 | false | none | false | false | false | false | false | false | 0 | none | 1 | positive | false | 0 | none | auto_resolved | high | Quick question: I have a CSV with user sessions. Show me a quick breakdown of sessions per country and device type. | I'll keep this concrete and measurable. Goal: handle token usage reliably, with guardrails for formatting and failure handling. Plan: - Use schema validation (e.g., Pydantic) to prevent bad inputs from propagating. - Start by... | 1,554.865424 | 2025-02-12 | 7 | 2 | false | false | 0.9685 | I have a CSV with user sessions. Show me a quick breakdown of sessions per country and device type. | You are a data analysis assistant helping users reason about data, metrics, and models. Explain trade-offs, outline steps, and keep the language accessible but precise. | true | 62646e5a7026d4f68ab1867a730451326aa3b664 | 228 | true |
INT000166 | S00566 | U00030 | 2025-04-07T16:08:57Z | train | web_app | data_analysis | IN | APAC | free | individual | support_strict | google | gemini-1.5-pro | 0.42 | 0.921 | 1,024 | 873 | 556 | 1,429 | 954 | 0.45728 | 0.665 | false | none | false | false | false | false | false | false | 3 | code_run|email_send|sql_query | null | no_feedback | false | 0 | none | auto_resolved | medium | Quick question: Help me compute the conversion rate per marketing channel and highlight the top 3 performers. Can you walk me through a good solution? | I'll keep this concrete and measurable. Goal: handle token usage reliably, with guardrails for formatting and failure handling. Plan: - Add telemetry fields for latency/tokens/cost so you can debug regressions... | 1,497.903564 | 2025-04-07 | 16 | 0 | false | true | 1.5701 | Help me compute the conversion rate per marketing channel and highlight the top 3 performers. | You are a data analysis assistant helping users reason about data, metrics, and models. Explain trade-offs, outline steps, and keep the language accessible but precise. | false | a15ac130f4f428f0efac738d5c68655674fccc6e | 212 | true |
INT000167 | S00401 | U00212 | 2025-03-07T04:39:30Z | test | web_app | data_analysis | US | AMER | free | individual | creative_writer | meta | llama-3.1-70b | 0.143 | 0.907 | 2,048 | 1,058 | 862 | 1,920 | 1,453 | 0.384 | 0.932 | false | none | false | false | false | false | false | false | 3 | http_request|knowledge_lookup|workflow_orchestrator | 1 | positive | false | 0 | none | auto_resolved | low | Quick question: Help me compute the conversion rate per marketing channel and highlight the top 3 performers. A short, clear answer would be ideal. | Here's what I'd implement in production. Goal: handle tool-call success rate reliably, with guardrails for formatting and failure handling. Plan: - Prefer simple baselines first; then iterate based on measured failure slices. - Start by writing a... | 1,321.403992 | 2025-03-07 | 4 | 4 | false | false | 1.2274 | Help me compute the conversion rate per marketing channel and highlight the top 3 performers. | You are a data analysis assistant helping users reason about data, metrics, and models. Explain trade-offs, outline steps, and keep the language accessible but precise. | true | 814335a175319858ca540d38bfcac020ec643c3f | 249 | true |
INT000168 | S00006 | U00153 | 2025-03-23T05:40:05Z | val | web_app | brainstorming | AE | EMEA | free | team | internal_policy_qa | meta | llama-3.1-8b | 0.34 | 0.88 | 512 | 360 | 165 | 525 | 962 | 0.03152 | 0.809 | false | none | false | false | false | false | false | false | 0 | none | 2 | positive | false | 0 | none | auto_resolved | high | Brainstorm taglines for a productivity app that helps engineers stay focused. I'd like a practical, engineering-focused answer. | Here's what I'd implement in production. Goal: handle LLMOps dashboards reliably, with guardrails for formatting and failure handling. Plan: - Use schema validation (e.g., Pydantic) to prevent bad inputs from... | 545.738046 | 2025-03-23 | 5 | 6 | true | false | 2.1818 | Brainstorm taglines for a productivity app that helps engineers stay focused. | You are a creative assistant helping users explore ideas and options. Suggest multiple alternatives, structure them clearly, and stay practical. | true | 64f5ffde54eb84c78cab2d2147174c299cd15237 | 211 | true |
INT000169 | S01085 | U00029 | 2025-02-11T10:04:18Z | train | web_app | data_analysis | FR | EMEA | free | enterprise_team | general_assistant | openai | gpt-4o-mini | 0.483 | 0.968 | 1,024 | 324 | 384 | 708 | 1,609 | 0.01416 | 0.412 | true | formatting_error | false | false | false | true | false | false | 3 | code_run|email_send|knowledge_lookup | -2 | negative | true | 0 | switch_model_family | auto_resolved | low | Quick question: I want a short summary of key metrics for this dataset: total users, active users, and average revenue per user. Please explain step by step. | Let's make the behavior deterministic and observable. Signal: formatting_error while working on latency p95. Quick diagnosis: Your output likely violated the contract. Enforce schema validation and retry-on-invalid.... | 440.02486 | 2025-02-11 | 10 | 1 | false | true | 0.8438 | I want a short summary of key metrics for this dataset: total users, active users, and average revenue per user. | You are a data analysis assistant helping users reason about data, metrics, and models. Explain trade-offs, outline steps, and keep the language accessible but precise. | true | 0af51e7e9a94eeacb4db5e80369428b0b639dc4e | 218 | true |
INT000170 | S00213 | U00266 | 2025-02-19T09:56:23Z | train | internal_tool | content_writing | AE | EMEA | free | enterprise_team | code_assistant_secure | meta | llama-3.1-8b | 0.723 | 0.869 | 768 | 25 | 33 | 58 | 654 | 0.00348 | 0.504 | true | formatting_error | false | false | false | true | false | false | 0 | none | -1 | negative | true | 2 | manual_post_edit | escalated_to_human | critical | Can you help me debug this? Follow-up: the previous reply did not fully solve the issue. Draft a friendly email announcing a small price increase to our existing customers. A short, clear answer would be ideal. | Let's break this down into actionable steps. Signal: formatting_error while working on A/B testing. Quick diagnosis: Fail closed: if invalid, don't pass downstream; retry with the validator error. Recommended... | 88.685015 | 2025-02-19 | 9 | 2 | false | true | 0.7576 | Follow-up: the previous reply did not fully solve the issue. Draft a friendly email announcing a small price increase to our existing customers. | You are a writing assistant helping users craft clear, engaging, and well-structured content. Keep the tone natural, readable, and aligned with the user's intent. | true | 20f1dad1d39f7b8df6faaa166fe78024f8d03fd6 | 211 | true |
INT000171 | S00702 | U00316 | 2025-04-15T16:11:42Z | train | mobile_app | brainstorming | US | AMER | enterprise | individual | general_assistant | google | gemini-1.5-pro | 0.635 | 0.928 | 256 | 887 | 256 | 1,143 | 1,601 | 0.36576 | 0.996 | false | none | false | false | false | false | false | false | 0 | none | null | no_feedback | false | 0 | none | auto_resolved | low | I'm working on something and need help: I need naming ideas for an internal tool that tracks experiments and model runs. Please explain step by step. | Let's break this down into actionable steps. Goal: handle developer productivity reliably, with guardrails for formatting and failure handling. Plan: - Start by writing a small function boundary for developer productivity and add tests before... | 713.928795 | 2025-04-15 | 16 | 1 | false | true | 3.4648 | I need naming ideas for an internal tool that tracks experiments and model runs. | You are a creative assistant helping users explore ideas and options. Suggest multiple alternatives, structure them clearly, and stay practical. | false | 62cac6cf36e4ae3f199823b7f2ca46aa5e4a0aa7 | 245 | true |
INT000172 | S01329 | U00232 | 2025-03-15T15:37:33Z | train | mobile_app | data_analysis | CA | AMER | pro | team | code_assistant_secure | meta | llama-3.1-8b | 0.286 | 0.894 | 1,024 | 444 | 895 | 1,339 | 882 | 0.08032 | 0.826 | false | none | false | false | false | false | false | false | 0 | none | 2 | positive | false | 0 | none | auto_resolved | medium | I'm stuck on this and need some guidance: Help me compute the conversion rate per marketing channel and highlight the top 3 performers. A short, clear answer would be ideal. | This is a common pattern; the fix is straightforward. Goal: handle safety flag rate reliably, with guardrails for formatting and failure handling. Plan: - Use schema validation (e.g., Pydantic) to prevent bad inputs from... | 1,518.14059 | 2025-03-15 | 15 | 5 | true | true | 0.4961 | Help me compute the conversion rate per marketing channel and highlight the top 3 performers. | You are a data analysis assistant helping users reason about data, metrics, and models. Explain trade-offs, outline steps, and keep the language accessible but precise. | true | 4e275efe35fdc513983e7d3f3ad32b50cc391969 | 223 | true |
INT000173 | S00469 | U00161 | 2025-04-07T10:51:11Z | test | api | data_analysis | FR | EMEA | free | individual | analytics_helper | meta | llama-3.1-8b | 0.108 | 0.895 | 1,024 | 694 | 539 | 1,233 | 1,375 | 0.074 | 0.98 | false | none | false | false | false | false | false | false | 0 | none | 1 | positive | false | 0 | none | auto_resolved | medium | Can you help me debug this? Can you suggest a few plots to explore churn risk in this dataset of subscriptions? A short, clear answer would be ideal. | Let's break this down into actionable steps. Goal: handle cost per request reliably, with guardrails for formatting and failure handling. Plan: - Use schema validation (e.g., Pydantic) to prevent bad inputs from propagating. - Prefer simple baselines... | 896.727273 | 2025-04-07 | 10 | 0 | false | true | 1.2876 | Can you suggest a few plots to explore churn risk in this dataset of subscriptions? | You are a data analysis assistant helping users reason about data, metrics, and models. Explain trade-offs, outline steps, and keep the language accessible but precise. | true | fbaac57cdb64395ff60fbf79b148e22caaff1431 | 253 | true |
INT000174 | S00939 | U00235 | 2025-04-22T12:47:27Z | train | api | content_writing | BR | AMER | free | individual | internal_policy_qa | google | gemini-1.5-pro | 0.883 | 0.865 | 2,048 | 155 | 610 | 765 | 831 | 0.2448 | 0.878 | false | none | false | false | false | false | false | false | 0 | none | 2 | positive | false | 0 | none | auto_resolved | medium | I'm stuck on this and need some guidance: Rewrite this paragraph in a clearer and more professional tone. | Here are the guardrails I'd put in place. Goal: handle observability metrics reliably, with guardrails for formatting and failure handling. Plan: - Start by writing a small function boundary for observability metrics and add tests before... | 920.577617 | 2025-04-22 | 12 | 1 | false | true | 0.2541 | Rewrite this paragraph in a clearer and more professional tone. | You are a writing assistant helping users craft clear, engaging, and well-structured content. Keep the tone natural, readable, and aligned with the user's intent. | true | db47a2ba00ac8f8120dcd2fd47d3648a8f594c66 | 240 | true |
INT000175 | S00550 | U00241 | 2025-02-07T21:27:17Z | train | web_app | content_writing | US | AMER | enterprise | enterprise_team | code_assistant_secure | openai | gpt-4.1 | 0.844 | 0.846 | 1,536 | 391 | 1,014 | 1,405 | 1,024 | 0.562 | 0.826 | false | none | false | false | false | false | false | false | 0 | none | null | no_feedback | false | 0 | none | auto_resolved | medium | I'm working on something and need help: Draft a friendly email announcing a small price increase to our existing customers. Please explain step by step. | Let's make the behavior deterministic and observable. Goal: handle observability metrics reliably, with guardrails for formatting and failure handling. Plan: - Prefer simple baselines first; then iterate based on measured failure slices. - Add telemetry... | 1,372.070313 | 2025-02-07 | 21 | 4 | false | false | 0.3856 | Draft a friendly email announcing a small price increase to our existing customers. | You are a writing assistant helping users craft clear, engaging, and well-structured content. Keep the tone natural, readable, and aligned with the user's intent. | false | 20f1dad1d39f7b8df6faaa166fe78024f8d03fd6 | 256 | true |
INT000176 | S00267 | U00201 | 2025-03-23T22:16:15Z | train | mobile_app | content_writing | US | AMER | pro | individual | analytics_helper | openai | gpt-4o-mini | 0.901 | 0.914 | 1,536 | 635 | 910 | 1,545 | 775 | 0.03088 | 0.785 | false | none | false | false | false | false | false | false | 0 | none | null | no_feedback | false | 0 | none | auto_resolved | critical | Quick question: Write a concise product description for a new wireless mouse aimed at software developers. Can you walk me through a good solution? | Let's break this down into actionable steps. Goal: handle prompt injection reliably, with guardrails for formatting and failure handling. Plan: - Prefer simple baselines first; then iterate based on measured failure slices. - Start by... | 1,993.548387 | 2025-03-23 | 22 | 6 | true | false | 0.6978 | Write a concise product description for a new wireless mouse aimed at software developers. | You are a writing assistant helping users craft clear, engaging, and well-structured content. Keep the tone natural, readable, and aligned with the user's intent. | false | 9716fa3b2c6fe39ac0f75ddaad781dfa757a0de9 | 237 | true |
INT000177 | S01561 | U00067 | 2025-03-11T01:42:45Z | train | internal_tool | content_writing | AU | APAC | pro | enterprise_team | support_strict | anthropic | claude-3.5-sonnet | 0.787 | 0.851 | 768 | 1,249 | 339 | 1,588 | 1,910 | 0.50816 | 0.411 | true | hallucination | true | false | false | false | false | false | 0 | none | null | no_feedback | false | 2 | switch_model_family | escalated_to_human | low | Quick question: Follow-up: the previous reply did not fully solve the issue. Rewrite this paragraph in a clearer and more professional tone. Please explain step by step. | We can make this robust with a few changes. Signal: hallucination while working on observability metrics. Quick diagnosis: Refuse when confidence is low; avoid confident guesses. Recommended... | 831.413613 | 2025-03-11 | 1 | 1 | false | false | 3.6844 | Follow-up: the previous reply did not fully solve the issue. Rewrite this paragraph in a clearer and more professional tone. | You are a writing assistant helping users craft clear, engaging, and well-structured content. Keep the tone natural, readable, and aligned with the user's intent. | false | 75b7366387b6cf4f5631ec9dd298a359e5bb9c1d | 193 | true |
INT000178 | S00884 | U00107 | 2025-02-01T05:39:47Z | test | api | content_writing | US | AMER | enterprise | enterprise_team | analytics_helper | openai | gpt-4o-mini | 0.915 | 0.943 | 768 | 1,877 | 335 | 2,212 | 1,073 | 0.04424 | 1 | false | none | false | false | false | false | false | false | 0 | none | null | no_feedback | false | 0 | none | escalated_to_human | low | Quick question: Write a concise product description for a new wireless mouse aimed at software developers. Can you walk me through a good solution? | Here are the guardrails I'd put in place. Goal: handle LLM reliability reliably, with guardrails for formatting and failure handling. Plan: - Prefer simple baselines first; then iterate based on measured failure slices. - Use schema validation... | 2,061.509786 | 2025-02-01 | 5 | 5 | true | false | 5.603 | Write a concise product description for a new wireless mouse aimed at software developers. | You are a writing assistant helping users craft clear, engaging, and well-structured content. Keep the tone natural, readable, and aligned with the user's intent. | false | 9716fa3b2c6fe39ac0f75ddaad781dfa757a0de9 | 246 | true |
INT000179 | S00815 | U00322 | 2025-02-16T08:48:19Z | train | web_app | brainstorming | AU | APAC | free | individual | creative_writer | openai | gpt-4o-mini | 0.716 | 0.984 | 1,024 | 232 | 198 | 430 | 2,639 | 0.0086 | 0.734 | false | none | false | false | false | false | false | false | 0 | none | 2 | positive | false | 0 | none | auto_resolved | low | I'm stuck on this and need some guidance: Give me ten creative ideas for a YouTube series about learning data science from scratch. | This is a common pattern; the fix is straightforward. Goal: handle knowledge base search reliably, with guardrails for formatting and failure handling. Plan: - Prefer simple baselines first; then iterate based on measured failure slices. - Add telemetry... | 162.940508 | 2025-02-16 | 8 | 6 | true | false | 1.1717 | Give me ten creative ideas for a YouTube series about learning data science from scratch. | You are a creative assistant helping users explore ideas and options. Suggest multiple alternatives, structure them clearly, and stay practical. | true | 6f52c3b4cdce68aa614a61f324617ae8d0e91690 | 256 | true |
INT000180 | S00907 | U00410 | 2025-02-24T15:18:54Z | train | mobile_app | data_analysis | AE | EMEA | free | enterprise_team | code_assistant_secure | openai | gpt-4o-mini | 0.384 | 0.95 | 1,024 | 303 | 466 | 769 | 3,549 | 0.0154 | 0.655 | false | none | false | false | false | false | false | false | 2 | email_send|workflow_orchestrator | 1 | positive | false | 0 | none | auto_resolved | low | I'm working on something and need help: I have a CSV with user sessions. Show me a quick breakdown of sessions per country and device type. A short, clear answer would be ideal. | Below is a decision-focused breakdown. Goal: handle latency p95 reliably, with guardrails for formatting and failure handling. Plan: - Add telemetry fields for latency/tokens/cost so you can debug regressions quickly. - Use schema validation... | 216.680755 | 2025-02-24 | 15 | 0 | false | true | 0.6502 | I have a CSV with user sessions. Show me a quick breakdown of sessions per country and device type. | You are a data analysis assistant helping users reason about data, metrics, and models. Explain trade-offs, outline steps, and keep the language accessible but precise. | true | 4e275efe35fdc513983e7d3f3ad32b50cc391969 | 244 | true |
INT000181 | S00121 | U00018 | 2025-02-19T16:02:59Z | train | mobile_app | brainstorming | GB | EMEA | free | enterprise_team | code_assistant_secure | openai | gpt-4.1 | 0.658 | 0.807 | 512 | 469 | 512 | 981 | 733 | 0.3924 | 0.744 | false | none | false | false | false | false | false | false | 0 | none | null | no_feedback | false | 0 | none | abandoned | medium | Quick question: Give me ten creative ideas for a YouTube series about learning data science from scratch. Please explain step by step. | Here's what I'd implement in production. Goal: handle knowledge base search reliably, with guardrails for formatting and failure handling. Plan: - Use schema validation (e.g., Pydantic) to prevent bad inputs from propagating. - Start by writing a... | 1,338.335607 | 2025-02-19 | 16 | 2 | false | true | 0.916 | Give me ten creative ideas for a YouTube series about learning data science from scratch. | You are a creative assistant helping users explore ideas and options. Suggest multiple alternatives, structure them clearly, and stay practical. | false | 9886b059d7fa547dc6fd91d4a5c80abdb317022d | 249 | true |
INT000182 | S00005 | U00224 | 2025-04-02T13:52:48Z | train | mobile_app | data_analysis | FR | EMEA | free | enterprise_team | analytics_helper | meta | llama-3.1-8b | 0.284 | 0.912 | 1,536 | 793 | 275 | 1,068 | 2,101 | 0.06408 | 0.814 | false | none | false | false | false | false | false | false | 0 | none | null | no_feedback | false | 0 | none | auto_resolved | low | Context for you: Help me compute the conversion rate per marketing channel and highlight the top 3 performers. Please explain step by step. | This is a common pattern; the fix is straightforward. Goal: handle latency p95 reliably, with guardrails for formatting and failure handling. Plan: - Use schema validation (e.g., Pydantic) to prevent bad inputs from propagating. - Start by... | 508.329367 | 2025-04-02 | 13 | 2 | false | true | 2.8836 | Help me compute the conversion rate per marketing channel and highlight the top 3 performers. | You are a data analysis assistant helping users reason about data, metrics, and models. Explain trade-offs, outline steps, and keep the language accessible but precise. | false | fbaac57cdb64395ff60fbf79b148e22caaff1431 | 242 | true |
INT000183 | S01382 | U00143 | 2025-04-01T16:32:25Z | test | web_app | brainstorming | EG | EMEA | free | team | code_assistant_secure | google | gemini-1.5-pro | 0.58 | 0.9 | 1,536 | 138 | 127 | 265 | 2,688 | 0.0848 | 0.593 | false | none | false | false | false | false | false | false | 0 | none | null | no_feedback | false | 0 | none | auto_resolved | low | Context for you: Brainstorm taglines for a productivity app that helps engineers stay focused. | Below is a practical plan you can execute. Goal: handle LLMOps dashboards reliably, with guardrails for formatting and failure handling. Plan: - Start by writing a small function boundary for LLMOps dashboards and add... | 98.58631 | 2025-04-01 | 16 | 1 | false | true | 1.0866 | Brainstorm taglines for a productivity app that helps engineers stay focused. | You are a creative assistant helping users explore ideas and options. Suggest multiple alternatives, structure them clearly, and stay practical. | false | 9886b059d7fa547dc6fd91d4a5c80abdb317022d | 220 | true |
INT000184 | S00538 | U00097 | 2025-03-19T06:21:26Z | train | web_app | coding_assistant | FR | EMEA | free | individual | code_assistant_secure | anthropic | claude-3.5-sonnet | 0.178 | 0.886 | 512 | 246 | 165 | 411 | 506 | 0.13152 | 0.942 | false | none | false | false | false | false | false | false | 0 | none | null | no_feedback | false | 0 | none | auto_resolved | high | Context for you: I need help understanding why this asynchronous code is not awaiting correctly. Explain with a small code snippet. A short, clear answer would be ideal. | Here are the guardrails I'd put in place. Goal: handle FastAPI endpoint design reliably, with guardrails for formatting and failure handling. Plan: - Use schema validation (e.g., Pydantic) to prevent bad inputs from... | 812.252964 | 2025-03-19 | 6 | 2 | false | false | 1.4909 | I need help understanding why this asynchronous code is not awaiting correctly. Explain with a small code snippet. | You are a senior software engineer helping users with code, debugging, and best practices. Provide clear explanations, structured steps, and high-quality code examples when relevant. | false | 52eadb52b3319d2dab569ce8fb653a0bee56a90f | 218 | true |
INT000185 | S01100 | U00360 | 2025-03-19T15:42:50Z | train | slack | customer_support | AE | EMEA | pro | individual | internal_policy_qa | local | custom-local-8b | 0.146 | 0.871 | 512 | 1,204 | 241 | 1,445 | 776 | 0.0058 | 0.77 | false | none | false | false | false | false | false | false | 0 | none | 2 | positive | false | 0 | none | escalated_to_human | low | Can you help me debug this? A user says they were double charged this month. Help me draft a clear explanation and next steps. Please explain step by step. | Let's turn this into a reliable workflow. Goal: handle account verification reliably, with guardrails for formatting and failure handling. Plan: - Use schema validation (e.g., Pydantic) to prevent bad inputs from propagating. - Start by writing a... | 1,862.113402 | 2025-03-19 | 15 | 2 | false | true | 4.9959 | A user says they were double charged this month. Help me draft a clear explanation and next steps. | You are a helpful customer support assistant for a SaaS product. Answer clearly, empathetically, and step-by-step while keeping the response concise and practical. | true | 03bf57d7adc83026d525c8192379aa203b71e1eb | 249 | true |
INT000186 | S01360 | U00223 | 2025-02-14T16:42:04Z | val | slack | customer_support | AE | EMEA | free | team | support_strict | openai | gpt-4o-mini | 0.922 | 0.859 | 256 | 515 | 256 | 771 | 1,521 | 0.0154 | 0.75 | false | none | false | false | false | false | false | false | 3 | code_run|http_request|workflow_orchestrator | -1 | negative | false | 0 | none | auto_resolved | low | A user says they were double charged this month. Help me draft a clear explanation and next steps. A short, clear answer would be ideal. | Here's what I'd implement in production. Goal: handle account verification reliably, with guardrails for formatting and failure handling. Plan: - Start by writing a small function boundary for account verification and add tests... | 506.903353 | 2025-02-14 | 16 | 4 | false | true | 2.0117 | A user says they were double charged this month. Help me draft a clear explanation and next steps. | You are a helpful customer support assistant for a SaaS product. Answer clearly, empathetically, and step-by-step while keeping the response concise and practical. | true | d029d0919f8e306d09ca7001d500c1e6f1c81007 | 230 | true |
INT000187 | S00105 | U00140 | 2025-04-28T07:13:13Z | train | web_app | brainstorming | IN | APAC | pro | individual | code_assistant_secure | meta | llama-3.1-8b | 0.631 | 0.988 | 256 | 183 | 77 | 260 | 674 | 0.0156 | 0.477 | true | toxicity | false | true | false | false | false | false | 0 | none | -2 | negative | false | 0 | manual_post_edit | escalated_to_human | medium | I'm working on something and need help: Brainstorm taglines for a productivity app that helps engineers stay focused. I'd like a practical, engineering-focused answer. | Here's the clean way to approach it. Signal: toxicity while working on customer support automation. Quick diagnosis: Apply safety filtering and tighten refusal rules; log safety flags for monitoring.... | 385.756677 | 2025-04-28 | 7 | 0 | false | false | 2.3766 | Brainstorm taglines for a productivity app that helps engineers stay focused. | You are a creative assistant helping users explore ideas and options. Suggest multiple alternatives, structure them clearly, and stay practical. | true | 9886b059d7fa547dc6fd91d4a5c80abdb317022d | 202 | true |
INT000188 | S00832 | U00379 | 2025-03-11T14:57:45Z | train | slack | data_analysis | BR | AMER | pro | individual | general_assistant | anthropic | claude-3.5-sonnet | 0.184 | 0.963 | 1,024 | 1,510 | 370 | 1,880 | 985 | 0.6016 | 0.51 | true | hallucination | true | false | false | false | false | false | 0 | none | null | no_feedback | false | 1 | regenerate_with_lower_temp | abandoned | low | Quick question: Follow-up: the previous reply did not fully solve the issue. I want a short summary of key metrics for this dataset: total users, active users, and average revenue per user. Please explain step by step. | Here's what I'd implement in production. Signal: hallucination while working on token usage. Quick diagnosis: Refuse when confidence is low; avoid confident guesses. Recommended actions: - Add... | 1,908.629442 | 2025-03-11 | 14 | 1 | false | true | 4.0811 | Follow-up: the previous reply did not fully solve the issue. I want a short summary of key metrics for this dataset: total users, active users, and average revenue per user. | You are a data analysis assistant helping users reason about data, metrics, and models. Explain trade-offs, outline steps, and keep the language accessible but precise. | false | 0af51e7e9a94eeacb4db5e80369428b0b639dc4e | 195 | true |
INT000189 | S00031 | U00069 | 2025-03-11T01:03:53Z | test | web_app | customer_support | IN | APAC | free | individual | code_assistant_secure | openai | gpt-4.1 | 0.769 | 0.952 | 768 | 130 | 341 | 471 | 3,504 | 0.1884 | 0.687 | false | none | false | false | false | false | false | false | 0 | none | null | no_feedback | false | 0 | none | abandoned | medium | Can you help me debug this? Hi, I'm having an issue with my billing. The invoice total doesn't match what I expected. Can you check it? I'd like a practical, engineering-focused answer. | Here are the guardrails I'd put in place. Goal: handle billing dispute reliably, with guardrails for formatting and failure handling. Plan: - Add telemetry fields for latency/tokens/cost so you can debug regressions quickly. - Start... | 134.417808 | 2025-03-11 | 1 | 1 | false | false | 0.3812 | Hi, I'm having an issue with my billing. The invoice total doesn't match what I expected. Can you check it? | You are a helpful customer support assistant for a SaaS product. Answer clearly, empathetically, and step-by-step while keeping the response concise and practical. | false | 1a8e13c1d6c8047debb9a230bf3ae08524ccb884 | 235 | true |
INT000190 | S00760 | U00130 | 2025-02-21T09:35:21Z | train | api | data_analysis | FR | EMEA | enterprise | team | analytics_helper | anthropic | claude-3.5-sonnet | 0.356 | 0.905 | 768 | 777 | 768 | 1,545 | 909 | 0.4944 | 0.672 | false | none | false | false | false | false | false | false | 3 | code_run|email_send|http_request | 2 | positive | false | 0 | none | auto_resolved | high | Can you help me debug this? Help me compute the conversion rate per marketing channel and highlight the top 3 performers. Please explain step by step. | Let's make the behavior deterministic and observable. Goal: handle latency p95 reliably, with guardrails for formatting and failure handling. Plan: - Prefer simple baselines first; then iterate based on measured failure slices. -... | 1,699.669967 | 2025-02-21 | 9 | 4 | false | true | 1.0117 | Help me compute the conversion rate per marketing channel and highlight the top 3 performers. | You are a data analysis assistant helping users reason about data, metrics, and models. Explain trade-offs, outline steps, and keep the language accessible but precise. | true | fbaac57cdb64395ff60fbf79b148e22caaff1431 | 232 | true |
INT000191 | S01446 | U00055 | 2025-04-10T07:49:05Z | train | internal_tool | brainstorming | GB | EMEA | enterprise | enterprise_team | code_assistant_secure | local | custom-local-8b | 0.391 | 0.899 | 2,048 | 587 | 755 | 1,342 | 1,035 | 0.00536 | 0.844 | false | none | false | false | false | false | false | false | 0 | none | null | no_feedback | false | 0 | none | auto_resolved | medium | I'm stuck on this and need some guidance: This is coming from an internal team in our company. I need naming ideas for an internal tool that tracks experiments and model runs. Can you walk me through a good solution? | Let's make the behavior deterministic and observable. Goal: handle cost optimization reliably, with guardrails for formatting and failure handling. Plan: - Add telemetry fields for latency/tokens/cost so you can debug regressions quickly. -... | 1,296.618357 | 2025-04-10 | 7 | 3 | false | false | 0.7775 | This is coming from an internal team in our company. I need naming ideas for an internal tool that tracks experiments and model runs. | You are a creative assistant helping users explore ideas and options. Suggest multiple alternatives, structure them clearly, and stay practical. | false | 9886b059d7fa547dc6fd91d4a5c80abdb317022d | 243 | true |
INT000192 | S00867 | U00001 | 2025-02-24T01:42:30Z | test | api | internal_qa | IN | APAC | free | individual | general_assistant | google | gemini-1.5-pro | 0.288 | 0.858 | 768 | 626 | 17 | 643 | 669 | 0.20576 | 0.564 | true | hallucination | true | false | false | false | false | false | 0 | none | null | no_feedback | false | 0 | add_constraints_to_prompt | escalated_to_human | low | Can you help me debug this? Summarise this internal security policy into three bullet points for non-technical employees. | Here are the guardrails I'd put in place. Signal: hallucination while working on errors. Quick diagnosis: Refuse when confidence is low; avoid confident guesses. Recommended actions: - Implement... | 961.136024 | 2025-02-24 | 1 | 0 | false | false | 36.8235 | Summarise this internal security policy into three bullet points for non-technical employees. | You are an internal knowledge assistant helping teammates navigate documentation, policies, and internal tools. Be precise, concise, and point to relevant information and next steps. | false | 7cd4f5f09ad56cc7a4cac3fea1f795d4e2ccc742 | 197 | true |
INT000193 | S01024 | U00074 | 2025-04-08T08:18:21Z | test | web_app | data_analysis | GB | EMEA | enterprise | team | internal_policy_qa | meta | llama-3.1-8b | 0.361 | 0.985 | 2,048 | 1,522 | 181 | 1,703 | 1,150 | 0.1022 | 0.996 | false | none | false | false | false | false | false | false | 2 | http_request|vector_search | null | no_feedback | false | 0 | none | auto_resolved | low | I'm working on something and need help: I want a short summary of key metrics for this dataset: total users, active users, and average revenue per user. Can you walk me through a good solution? | Let's turn this into a reliable workflow. Goal: handle safety flag rate reliably, with guardrails for formatting and failure handling. Plan: - Prefer simple baselines first; then iterate based on measured failure... | 1,480.869565 | 2025-04-08 | 8 | 1 | false | false | 8.4088 | I want a short summary of key metrics for this dataset: total users, active users, and average revenue per user. | You are a data analysis assistant helping users reason about data, metrics, and models. Explain trade-offs, outline steps, and keep the language accessible but precise. | false | 62646e5a7026d4f68ab1867a730451326aa3b664 | 215 | true |
INT000194 | S01141 | U00311 | 2025-04-23T13:09:30Z | train | web_app | customer_support | GB | EMEA | pro | individual | support_strict | meta | llama-3.1-70b | 0.826 | 0.866 | 1,024 | 1,106 | 56 | 1,162 | 1,051 | 0.2324 | 0.699 | false | none | false | false | false | false | false | false | 0 | none | null | no_feedback | false | 0 | none | auto_resolved | low | I need a short, friendly reply to a customer who is asking about the status of their refund. A short, clear answer would be ideal. | Let's break this down into actionable steps. Goal: handle unexpected charge reliably, with guardrails for formatting and failure handling. Plan: - Start by writing a small function boundary for unexpected charge and add tests before optimizing. -... | 1,105.613701 | 2025-04-23 | 13 | 2 | false | true | 19.75 | I need a short, friendly reply to a customer who is asking about the status of their refund. | You are a helpful customer support assistant for a SaaS product. Answer clearly, empathetically, and step-by-step while keeping the response concise and practical. | false | d029d0919f8e306d09ca7001d500c1e6f1c81007 | 249 | true |
INT000195 | S00916 | U00271 | 2025-03-26T08:08:19Z | train | api | data_analysis | FR | EMEA | free | individual | creative_writer | meta | llama-3.1-70b | 0.192 | 0.855 | 1,024 | 643 | 18 | 661 | 660 | 0.1322 | 0.398 | true | safety_block | false | false | true | false | false | false | 3 | code_run|sql_query|workflow_orchestrator | -1 | negative | true | 0 | enable_tools | auto_resolved | low | Context for you: I want a short summary of key metrics for this dataset: total users, active users, and average revenue per user. | Below is a practical plan you can execute. Signal: safety_block while working on error rate. Quick diagnosis: Refuse safely and offer a legitimate alternative or a high-level explanation.... | 1,001.515152 | 2025-03-26 | 8 | 2 | false | false | 35.7222 | I want a short summary of key metrics for this dataset: total users, active users, and average revenue per user. | You are a data analysis assistant helping users reason about data, metrics, and models. Explain trade-offs, outline steps, and keep the language accessible but precise. | true | 814335a175319858ca540d38bfcac020ec643c3f | 190 | true |
INT000196 | S01226 | U00433 | 2025-04-03T02:55:14Z | test | web_app | customer_support | CA | AMER | free | individual | analytics_helper | anthropic | claude-3.5-sonnet | 0.454 | 0.834 | 2,048 | 792 | 505 | 1,297 | 983 | 0.41504 | 0.432 | true | safety_block | false | false | true | false | false | false | 0 | none | null | no_feedback | true | 1 | regenerate_with_lower_temp | auto_resolved | high | Can you help me debug this? Follow-up: the previous reply did not fully solve the issue. I need a short, friendly reply to a customer who is asking about the status of their refund. I'd like a practical, engineering-focused answer. | Below is a decision-focused breakdown. Signal: safety_block while working on account verification. Quick diagnosis: Log the trigger category to tune your safety policy. Recommended actions: - Add... | 1,319.430315 | 2025-04-03 | 2 | 3 | false | false | 1.5683 | Follow-up: the previous reply did not fully solve the issue. I need a short, friendly reply to a customer who is asking about the status of their refund. | You are a helpful customer support assistant for a SaaS product. Answer clearly, empathetically, and step-by-step while keeping the response concise and practical. | false | 598eba0ae15195966a9835b68c542f0b34fb88e4 | 198 | true |
INT000197 | S00177 | U00072 | 2025-04-16T15:03:59Z | val | web_app | data_analysis | FR | EMEA | enterprise | enterprise_team | creative_writer | meta | llama-3.1-8b | 0.172 | 0.875 | 2,048 | 802 | 15 | 817 | 2,161 | 0.04904 | 0.737 | false | none | false | false | false | false | false | false | 0 | none | null | no_feedback | false | 0 | none | escalated_to_human | low | I'm working on something and need help: I want a short summary of key metrics for this dataset: total users, active users, and average revenue per user. Can you walk me through a good solution? | We can make this robust with a few changes. Goal: handle error rate reliably, with guardrails for formatting and failure handling. Plan: - Add telemetry fields for latency/tokens/cost so you can debug regressions quickly. - Use schema validation... | 378.06571 | 2025-04-16 | 15 | 2 | false | true | 53.4667 | I want a short summary of key metrics for this dataset: total users, active users, and average revenue per user. | You are a data analysis assistant helping users reason about data, metrics, and models. Explain trade-offs, outline steps, and keep the language accessible but precise. | false | 814335a175319858ca540d38bfcac020ec643c3f | 248 | true |
INT000198 | S00702 | U00316 | 2025-04-15T16:22:32Z | train | mobile_app | brainstorming | US | AMER | enterprise | individual | support_strict | anthropic | claude-3.5-sonnet | 0.735 | 0.915 | 1,536 | 840 | 696 | 1,536 | 367 | 0.49152 | 0.813 | false | none | false | false | false | false | false | false | 0 | none | 2 | positive | false | 0 | none | auto_resolved | low | I'm stuck on this and need some guidance: I need naming ideas for an internal tool that tracks experiments and model runs. Please explain step by step. | We can make this robust with a few changes. Goal: handle developer productivity reliably, with guardrails for formatting and failure handling. Plan: - Use schema validation (e.g., Pydantic) to prevent bad inputs from... | 4,185.286104 | 2025-04-15 | 16 | 1 | false | true | 1.2069 | I need naming ideas for an internal tool that tracks experiments and model runs. | You are a creative assistant helping users explore ideas and options. Suggest multiple alternatives, structure them clearly, and stay practical. | true | e3da6c15e87a6fab71f2d2b735988233f5d1f30b | 219 | true |
INT000199 | S00007 | U00094 | 2025-03-22T09:34:57Z | train | mobile_app | customer_support | IN | APAC | pro | individual | analytics_helper | openai | gpt-4.1 | 0.478 | 0.863 | 1,024 | 1,109 | 499 | 1,608 | 1,443 | 0.6432 | 0.681 | false | none | false | false | false | false | false | false | 1 | email_send | 2 | positive | false | 0 | none | abandoned | low | I'm working on something and need help: Our customer is getting an error when trying to log in to their account. Can you help me debug this? Can you walk me through a good solution? | I'll keep this concrete and measurable. Goal: handle billing dispute reliably, with guardrails for formatting and failure handling. Plan: - Start by writing a small function boundary for billing dispute and add tests before optimizing. - Prefer... | 1,114.345114 | 2025-03-22 | 9 | 5 | true | true | 2.2224 | Our customer is getting an error when trying to log in to their account. Can you help me debug this? | You are a helpful customer support assistant for a SaaS product. Answer clearly, empathetically, and step-by-step while keeping the response concise and practical. | true | 598eba0ae15195966a9835b68c542f0b34fb88e4 | 247 | true |
INT000200 | S01133 | U00168 | 2025-02-07T21:24:11Z | train | api | data_analysis | US | AMER | free | team | general_assistant | local | custom-local-8b | 0.137 | 0.878 | 1,024 | 500 | 619 | 1,119 | 2,121 | 0.00448 | 0.98 | false | none | false | false | false | false | false | false | 0 | none | 2 | positive | false | 0 | none | auto_resolved | high | Help me compute the conversion rate per marketing channel and highlight the top 3 performers. A short, clear answer would be ideal. | We can make this robust with a few changes. Goal: handle cost per request reliably, with guardrails for formatting and failure handling. Plan: - Prefer simple baselines first; then iterate based on measured failure slices. - Use schema validation... | 527.58133 | 2025-02-07 | 21 | 4 | false | false | 0.8078 | Help me compute the conversion rate per marketing channel and highlight the top 3 performers. | You are a data analysis assistant helping users reason about data, metrics, and models. Explain trade-offs, outline steps, and keep the language accessible but precise. | true | 0af51e7e9a94eeacb4db5e80369428b0b639dc4e | 249 | true |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.