Data and AI struggles

2,149 problems · 379 companies

LLM apps

MLflow fails to return valid structured output

5 posts from 4 people4d activeRising, +4 this week

Posts per day

Posts per day5 posts, Nov 10 to Nov 22

The posts behind it

GitHub issuesmlflow/mlflow · atr-ipNov 20

[FR] Scoring server: emit valid JSON (`null`) instead of bare `NaN`/`Infinity` in prediction responses

...default `allow_nan=True`. These tokens are not JSON (RFC 8259), so strict clients fail to parse the whole response. Proposal: write non-finite values as `null` by default (or behind an opt-in setting such as `MLFLOW_SCORING_SERVER_STRICT_JSON=true`), and document the behaviour....

1 replies
GitHub issuesmlflow/mlflow · atr-ipNov 20

[BUG] Input examples and serving input examples contain bare `NaN` (invalid JSON) for missing values in float columns

### Issues Policy acknowledgement - [x] I have read and agree to submit bug reports in accordance with the issues policy ### Where did you encounter this bug? Local machine ### MLflow version - Client: 3.18.0 (also checked: `_handle_dataframe_nans` is unchanged on master) ###...

2 replies
GitHub issuesmlflow/mlflow · cravaniNov 19

Safety judge verdict can contradict its own rationale

...in evaluations and need verdicts that match the reasoning. > #### Why is it currently difficult to achieve this use case? `Safety` and `is_safe` have no way to change the output field order. The safety prompt already asks for the rationale first, but the default schema overrides...

1 replies
GitHub issuesmlflow/mlflow · Yut0MiuraNov 18

[FR]: Add JSON Lines output with automatic pagination to `mlflow traces search`

...introducing additional pagination logic in every consumer. > #### Why is it currently difficult to achieve this use case? `mlflow traces search` already supports filtering, ordering, field selection and pagination through continuation tokens. However, each invocation returns...

2 replies
GitHub issuesmlflow/mlflow · affan-DevOpNov 10

[FR] Handle `{{ }}` placeholders in `tool_calls` arguments consistently in chat prompts

...prompt registry and don't use this in production yet. > #### Why is it currently difficult to achieve this use case? `format()` returns `tool_calls` unchanged, so the only workaround is to fill the arguments yourself afterwards. `arguments` is a JSON-encoded string, so plain...

2 replies

Companies and products named

5 posts
2 posts
1 post
1 post
About this problem

Evidence

5 posts from 4 people in 1 place, about 3 a week over 11 days. Mostly on mlflow/mlflow. Tools named alongside: pandas, GitHub, Infinity, Databricks.

Frustration Frustration 0 of 3· Seen on GitHub issues

Summary, model-written

MLflow users experience errors when non-finite values break strict JSON response parsing, tool call arguments are returned unchanged, and trace searches demand manual consumer pagination logic.

How it was grouped

Posts that state a pain and match the "structured-output" rule about MLflow. First post Nov 10, 2026, latest Nov 20, 2026. Corroborated: 3 or more posts from 2 or more people or places. Method

The same problem elsewhere

Models fail to return valid structured output

19 posts from 17 people on GitHub issues.

Claude fails to return valid structured output

15 posts from 12 people on GitHub issues.

OpenAI fails to return valid structured output

12 posts from 12 people on GitHub issues, Hacker News.

Anthropic fails to return valid structured output

6 posts from 6 people on GitHub issues.

Codex fails to return valid structured output

6 posts from 3 people on GitHub issues.

LlamaIndex fails to return valid structured output

5 posts from 5 people on GitHub issues.

Model Context Protocol fails to return valid structured output

5 posts from 5 people on GitHub issues, Reddit.

vLLM fails to return valid structured output

5 posts from 5 people on GitHub issues.

LangChain fails to return valid structured output

5 posts from 4 people on GitHub issues.

Ollama fails to return valid structured output

4 posts from 3 people on GitHub issues.

Gemini fails to return valid structured output

4 posts from 4 people on GitHub issues, Reddit.

ChatGPT fails to return valid structured output

3 posts from 3 people on Hacker News, Stack Overflow, GitHub issues.

Other MLflow problems

MLflow upgrades break working setups

21 posts from 17 people on GitHub issues.

MLflow users keep filing bugs

Top post: "Prompt Registry allows modification of immutable prompt templates via reserved version tags", one of 17 posts from 12 people on GitHub issues.

MLflow error messages are cryptic

5 posts from 4 people on GitHub issues.

Serving models on MLflow is harder than it should be

17 posts from 16 people on GitHub issues.

Trouble with MLflow trace

Top post: "TypeScript tracing drops every trace whose root span has a remote parent (incoming traceparent)", one of 11 posts from 10 people on GitHub issues.

Trouble with MLflow list

Top post: "`users/list` returns users from other workspaces when workspaces are enabled", one of 7 posts from 6 people on GitHub issues.

MLflow rate limits and usage caps block work

6 posts from 5 people on GitHub issues.

Trouble with MLflow traces

Top post: "Add a typed client API to submit Issue Detection for traces", one of 6 posts from 4 people on GitHub issues.

Trouble with MLflow server

Top post: "Test requirements: test_http_artifact_root_uses_presigned_only_split_server 403 recurs after #27212 was closed", one of 4 posts from 3 people on GitHub issues.

Trouble with MLflow evaluation

Top post: "Cancel button in the evaluation dataset Add record panel does nothing", one of 4 posts from 3 people on GitHub issues.

MLflow vulnerabilities are hard to triage and patch

4 posts from 4 people on Glyph Ibis.

Experiments on MLflow are hard to reproduce and track

4 posts from 4 people on GitHub issues.

History

Summary written by a model: MLflow users experience errors when non-finite values break strict JSON response parsing, tool call arguments are returned unchanged, and trace searches demand manual consumer pagination logic.

Nov 22, 2026

First seen with 5 posts.

Nov 22, 2026