LangChain upgrades break working setups
11 posts from 11 people11d activeFading, +0 this week
Posts per day
The posts behind it
langchain-qdrant: add_texts silently stores nothing when texts is a generator
### Submission checklist - [x] This is a bug, not a usage question. - [x] I added a clear and descriptive title that summarizes this issue. - [x] I used the GitHub search to find a similar question and didn't find it. - [x] I am sure that this is a bug in LangChain rather than...
`ModelCallLimitMiddleware` does not count a model call when a later-listed `after_model` hook jumps away, so `run_limit` is not a bound
...in [Chemclaw3](https://github.com/8fqycwdt8v-oss/Chemclaw3); the first-party workaround named above is what we run today.
LLM cache stores a generation that fails structured-output parsing and then replays it
...### Related Issues / PRs This is not #29003 / #27264 / #25613. Those are about failing to deserialize a **valid** structured result from SQLite, Redis, or Elasticsearch. Here the generation is stored and read back fine. The stored generation is one the structured-output parser...
I built an open-source tool for replaying and debugging AI agent executions. Looking for contributors and feedback.
...executions to investigate failures.. Fork executions to experiment with fixes.. Minimize failing traces to make bugs easier to isolate.. Generate regression tests to help prevent bugs from returning.. The project has reached v0.2.1 , and I've been working on testing,...
[Bug]: FunctionTool returns an unawaited coroutine for async callable instances
### Bug Description Passing a service instance with an async `__call__` to `FunctionTool.from_defaults(fn=...)` does not execute it. Both `call()` and `acall()` return a successful `ToolOutput` containing an unawaited coroutine. Explicit name and description are provided. ###...
core: `RouterRunnable` batch methods ignore `return_exceptions` for unknown routes
### Submission checklist - [x] This is a bug in `langchain-core`, not a usage question. - [x] I searched GitHub issues and pull requests and found no duplicate of this runtime behavior. - [x] The bug reproduces on the latest stable release (`langchain-core 1.6.12`) and current...
V3 streaming drops provider IDs from text and reasoning block-start events
When a provider supplies an ID for a text or reasoning block, V3 drops that ID from the block-start event, even though it remains in the completed message. This makes it harder for streaming consumers to match the incoming block with its final version. The proposed fix preserves...
extract_sub_links produces malformed URL for query-only relative links when page URL has a query
### Submission checklist - [x] This is a bug, not a usage question. - [x] I added a clear and descriptive title that summarizes this issue. - [x] I used the GitHub search to find a similar question and didn't find it. - [x] I am sure that this is a bug in LangChain rather than...
v3 streaming: tool-call `block-delta` events are snapshots, so their size grows quadratically with the tool-call arguments
### Submission checklist - [x] This is a bug, not a usage question. - [x] I added a clear and descriptive title that summarizes this issue. - [x] I used the GitHub search to find a similar question and didn't find it. - [x] I am sure that this is a bug in LangChain rather than...
`with_fallbacks(exception_key=...)` writes the exception into the caller's own input mapping Package:langchain-core
### Submission checklist - [x] This is a bug, not a usage question. - [x] I added a clear and descriptive title that summarizes this issue. - [x] I used the GitHub search to find a similar question and didn't find it. - [x] I am sure that this is a bug in LangChain rather than...
CriteriaResultOutputParser takes the last or first letter of a word as the Y/N verdict (`match_last`/`match_first` have no `\b`)
...word boundary, and none of the cuts contains a verdict line: - A 6-sentence reply for a failing submission has 49 cuts. 3 score 1 (for example `...The submission states that the capital of Germany`, value `y`), 4 score 0 and 42 score None. - A reply for a passing submission has...
Companies and products named
About this problem
Evidence
11 posts from 11 people in 3 places, about 2 a week over 49 days. Mostly on langchain-ai/langchain, run-llama/llama_index, r/LangChain. Tools named alongside: GitHub, Anthropic, OpenAI, Perplexity.
Frustration Frustration 0 of 3· Seen on GitHub issues, Reddit
How it was grouped
Posts that state a pain and match the "breaking-changes" rule about LangChain. First post Sep 24, 2026, latest Nov 11, 2026. Corroborated: 3 or more posts from 2 or more people or places. Method
The same problem elsewhere
GitHub upgrades break working setups
109 posts from 94 people on GitHub issues, Hacker News, Stack Overflow.
Upgrades and new releases break working setups
94 posts from 76 people on GitHub issues, Reddit, App Store reviews.
Claude Code upgrades break working setups
82 posts from 73 people on GitHub issues, Hacker News, Reddit.
Codex upgrades break working setups
72 posts from 65 people on GitHub issues, Reddit.
DuckDB upgrades break working setups
55 posts from 42 people on GitHub issues.
Claude upgrades break working setups
46 posts from 44 people on GitHub issues, Reddit, Hacker News.
Ollama upgrades break working setups
40 posts from 20 people on GitHub issues.
ClickHouse upgrades break working setups
37 posts from 21 people on GitHub issues.
OpenAI upgrades break working setups
32 posts from 32 people on GitHub issues, Reddit, Community forums.
Anthropic upgrades break working setups
30 posts from 24 people on GitHub issues.
ChatGPT upgrades break working setups
29 posts from 29 people on GitHub issues, Community forums, App Store reviews.
PyTorch upgrades break working setups
29 posts from 23 people on GitHub issues, Reddit.
Other LangChain problems
LangChain fails to return valid structured output
5 posts from 4 people on GitHub issues.
LangChain users keep filing bugs
Top post: "langchain 1.4.0: any after_model middleware using documented `jump_to=tools` bypasses the HITL", one of 27 posts from 24 people on GitHub issues, Reddit.
Trouble with LangChain agent
Top post: "langchain-classic: importing a moved agent from `agent_toolkits` raises `ValueError` instead of `ImportError`", one of 8 posts from 7 people on GitHub issues, Reddit, Stack Overflow.
RAG on LangChain returns irrelevant or missing context
8 posts from 7 people on Reddit, GitHub issues, Stack Overflow.
LangChain agents loop, stall or fail midway
6 posts from 6 people on GitHub issues, Reddit.
LangChain rate limits and usage caps block work
4 posts from 3 people on GitHub issues.
Trouble with LangChain core
Top post: "core: check_valid_template raises ValueError for mustache template_format", one of 4 posts from 3 people on GitHub issues.
Trouble with LangChain metadata
Top post: "xai: `ChatXAI` drops LangSmith gateway metadata when streaming", one of 4 posts from 4 people on GitHub issues.
Trouble with LangChain agents
Top post: "Is it normal for a LangGraph supervisor to consume as many tokens as all 8 agents combined? How can we", one of 4 posts from 4 people on Reddit.
Trouble with LangChain test
Top post: "langchain: MCP adapter test asserts a POSIX-only error message, fails on Windows where the drive letter", one of 3 posts from 3 people on GitHub issues.
LangChain error messages are cryptic
3 posts from 2 people on GitHub issues.
LangChain vulnerabilities are hard to triage and patch
3 posts from 1 person on GitHub issues. Unverified.
History