AIOpsAnalyst

It is still running

Every six hours a scheduled job reads the newest issues from huggingface/transformers, vercel/next.js, langchain-ai/langchain, groups and labels them the same way, runs the same rulebook, and commits the result. It decides but files nothing: below is what the rulebook would do with each issue. The numbers elsewhere on this site come from a frozen corpus so they stay checkable; this page is the part that moves.

  • 48Would become a ticket
  • 0Wakes a person
  • 11Drafted, waits for approval
  • 1Routine, dropped
  • 0Not sure, recorded

Last run , 60 issues. Every run is in the commit history.

  1. ShellSession.execute: Command include "exit N" can lead to a TimeOut

    langchain-ai/langchain ·

    ShellSession.execute may timeout when a command includes "exit N" causing unexpected delays

    errormedium0.73 sure Drafted, waits for approval

  2. Support forwarding MCP _meta on tool calls and surfacing server response metadata

    langchain-ai/langchain ·

    Add opt‑in support to forward and expose MCP _meta metadata on tool calls and responses

    feature requestlow0.97 sure Would become a ticket

  3. RecursiveCharacterTextSplitter with regex separator raises error if there is only 1 match of the regex

    langchain-ai/langchain ·

    RecursiveCharacterTextSplitter throws an exception when using a regex separator that matches only once

    errorlow0.92 sure Would become a ticket

  4. Deterministic pre-execution tool-call guard (operator-declared policy) — working pattern + gate test

    langchain-ai/langchain ·

    Request to add a deterministic pre‑execution tool‑call guard for LangChain tools with policy checks and audit ledger

    feature requestlow0.97 sure Would become a ticket

  5. [BUG] MLukeTokenizer emits `<s>` instead of `<ent>` on all entity tasks

    huggingface/transformers ·

    MLukeTokenizer outputs <s> token instead of <ent> for all entity tasks, breaking entity labeling

    errormedium0.97 sure Would become a ticket

  6. Turbopack builds fail for edge route handlers when cacheHandlers is set

    vercel/next.js ·

    Turbopack builds fail for edge route handlers when cacheHandlers is enabled, while webpack builds succeed

    errormedium0.96 sure Would become a ticket

  7. Turbopack dev server deadlocks deferred route compilation after a proxy request (16.4.0)

    vercel/next.js ·

    Turbopack dev server hangs during deferred route compilation after a proxy request in Next.js 16.4.0

    errormedium0.90 sure Would become a ticket

  8. UnboundLocalError in SuperPointForKeypointDetection.forward under torch.compile

    huggingface/transformers ·

    UnboundLocalError occurs when compiling SuperPoint model forward function with torch.compile

    errormedium0.97 sure Would become a ticket

  9. AutoVideoProcessor fails to load backend-specific remote auto_map entries

    huggingface/transformers ·

    AutoVideoProcessor fails to load backend-specific remote auto_map entries when using trust_remote_code

    errormedium0.90 sure Would become a ticket

  10. [Error Overlay] Dev server aborts: next-code-frame panic on multi-byte source (terminal width)

    vercel/next.js ·

    Next.js dev server crashes due to a panic in next-code-frame when terminal width changes with multibyte source files

    crashmedium0.95 sure Would become a ticket

  11. next.config.ts: relative imports resolve against process.cwd() instead of the config file's directory

    vercel/next.js ·

    Relative imports in next.config.ts resolve against process.cwd() instead of the config file's directory, causing build issues

    errormedium0.96 sure Would become a ticket

  12. Cache Components enabled causes HTTP 500 / Error 1101 on Cloudflare Workers

    vercel/next.js ·

    Enabling Cache Components in Next.js on Cloudflare Workers causes HTTP 500 / Error 1101 failures

    errormedium0.92 sure Would become a ticket

  13. Docs: Preserving UI State guide missing warning for imperative DOM/WebGL libraries (Mapbox GL, Three.js, etc.)

    vercel/next.js ·

    Docs missing warning for imperative DOM/WebGL libraries in Preserving UI State guide

    feature requestlow0.97 sure Would become a ticket

  14. pipeline(..., model_kwargs={"code_revision": ...}) raises TypeError, and top-level code_revision does not reach the model load

    huggingface/transformers ·

    pipeline raises TypeError when using model_kwargs code_revision, preventing model load

    errormedium0.94 sure Would become a ticket

  15. core: expose the prompt template to chat model and LLM callbacks so tracers can attach it to GenAI spans

    langchain-ai/langchain ·

    Request to expose prompt template to chat model callbacks for tracing GenAI spans

    feature requestlow0.97 sure Would become a ticket

  16. HunYuan-VL beam search gives returned sequences the wrong image for batched prompts

    huggingface/transformers ·

    Beam search returns incorrect images for batched prompts in HunYuan-VL model

    errormedium0.96 sure Would become a ticket

  17. Parallel route slot matched by [...catchAll] keeps the old root param (locale) after client navigation in production

    vercel/next.js ·

    Parallel route slot retains old locale param after client navigation, causing incorrect routing behavior

    errormedium0.92 sure Would become a ticket

  18. Torch fallbacks of the linear-attention ops drop `seq_idx` / `cu_seqlens`, so packed documents leak without the optional packages

    huggingface/transformers ·

    Torch fallback for linear-attention ops drops seq_idx/cu_seqlens causing packed document memory leaks when optional packages are missing

    errormedium0.93 sure Would become a ticket

  19. SDPA: enable_gqa=True silently falls back to the math kernel at batch 1 on torch < 2.14 without Flash Attention

    huggingface/transformers ·

    SDPA with enable_gqa=True falls back to slower math kernel on torch <2.14, losing Flash Attention

    errormedium0.92 sure Would become a ticket

  20. Turbopack: panic when a PostCSS plugin prints a stack trace while it loads

    vercel/next.js ·

    Turbopack panics when a PostCSS plugin outputs a stack trace during loading, causing the dev server to crash

    crashmedium0.90 sure Would become a ticket

  21. Chameleon image generation low quality.

    huggingface/transformers ·

    User reports low-quality images from model.generate and asks for a conversion script/help.

    questionlow0.90 sure Routine, dropped

  22. Support fraction-of-context-window option for trim_tokens_to_summarize in SummarizationMiddleware

    langchain-ai/langchain ·

    Request to add fraction-of-context-window option to trim_tokens_to_summarize in SummarizationMiddleware

    feature requestlow0.97 sure Would become a ticket

  23. `experimental.turbopackRustReactCompiler` drops development options the Babel React Compiler path sets (Fast Refresh memo-cache reset, anonymous function names, flow suppressions, TS source type)

    vercel/next.js ·

    Experimental turbopack Rust React Compiler omits development options set by Babel React Compiler, causing missing features like Fast Refresh

    errormedium0.73 sure Drafted, waits for approval

  24. langchain-typesafe: NaN/Infinity in state or criteria crash invoke with uncaught ValueError from httpx2 internals

    langchain-ai/langchain ·

    Langchain-typesafe crashes with uncaught ValueError due to NaN/Infinity in state or criteria

    crashhigh0.94 sure Would become a ticket

  25. Native TP does not shard qwen3_vl_moe experts (weights fully replicated per rank)

    huggingface/transformers ·

    Native tensor parallel fails to shard qwen3_vl_mo e experts, replicating weights across ranks instead of sharding

    errormedium0.92 sure Would become a ticket

  26. AutoTokenizer produces wrong token IDs for OLMo2, HyperClovaX, DeepSeek-R1-Distill-Llama, Yi, and others (v5 regression)

    huggingface/transformers ·

    AutoTokenizer returns incorrect token IDs for several models, indicating a regression in the library

    errormedium0.97 sure Would become a ticket

  27. Gemma 3/4 multimodal models fail with tp_plan="auto" because of mixed Tensor/DTensor usage

    huggingface/transformers ·

    Gemma 3/4 multimodal models crash when using tp_plan="auto" due to mixed Tensor/DTensor usage in lm_head

    errormedium0.92 sure Would become a ticket

  28. LayoutLM rejects SDPA despite using the unified attention interface

    huggingface/transformers ·

    LayoutLM model fails to use SDPA because it lacks the _supports_sdpa flag despite unified attention interface

    errormedium0.92 sure Would become a ticket

  29. [serge] integration failure triage - 2026-10-11

    huggingface/transformers ·

    Automated CI integration failures reported for the daily window, requiring investigation

    errorhigh0.78 sure Drafted, waits for approval

  30. Prefetch loop: optimistic routing drops a parallel slot's catch-all param next to a static segment

    vercel/next.js ·

    Prefetch loop caused by optimistic routing dropping a catch-all param leads to unexpected extra requests

    errormedium0.78 sure Drafted, waits for approval

  31. Whisper large-v3 and large-v3-turbo: config declares float16 but WhisperFeatureExtractor emits float32, so the manual path raises

    huggingface/transformers ·

    Whisper large-v3 models raise a type mismatch error due to config float16 vs feature extractor float32

    errormedium0.97 sure Would become a ticket

  32. A `text-plain` content block with inline `text` is rejected by the Bedrock, Gemini and OpenAI converters

    langchain-ai/langchain ·

    Plain text content blocks are rejected by Bedrock, Gemini, and OpenAI converters, causing message conversion failures.

    errormedium0.92 sure Would become a ticket

  33. Vercel: launcher crashes with Cannot find module './.next/server/pages/_next/image.js' when an image optimizer local source returns 5xx

    vercel/next.js ·

    Next.js launcher crashes with a module not found error when the image optimizer source returns a 5xx status code.

    crashhigh0.95 sure Would become a ticket

  34. Mamba2: non-fused gated RMSNorm ignores n_groups and disagrees with fused training

    huggingface/transformers ·

    Mamba2 non-fused gated RMSNorm implementation differs from fused version when n_groups>1, causing incorrect outputs and gradients

    errormedium0.94 sure Would become a ticket

  35. Turbopack: webpackChunkName ignored when migrating from webpack

    vercel/next.js ·

    Chunk name from webpackChunkName is ignored when switching to Turbopack, causing missing chunk naming

    errorlow0.90 sure Would become a ticket

  36. Turbopack dev: next/dynamic in Pages Router _app preloads a 404 chunk (regression in 16.5.0-canary.8)

    vercel/next.js ·

    Turbopack dev preloads a missing 404 chunk in next/dynamic Pages Router, regression in 16.5.0-canary.8

    errormedium0.85 sure Drafted, waits for approval

  37. Gemma 4 `response_template` parses reasoning into `thinking`, which its chat template does not read back

    huggingface/transformers ·

    Gemma 4 response_template puts reasoning in a 'thinking' field that the chat template doesn't consume, causing parsing issues

    errormedium0.94 sure Would become a ticket

  38. Turbopack dev: next/dynamic preloads a chunk that is never emitted (404) when the component has an import()

    vercel/next.js ·

    Turbopack dev server preloads a non‑existent chunk for next/dynamic causing 404 errors

    errormedium0.93 sure Would become a ticket

  39. Chunked prefill mishandles prefix-cache continuation

    huggingface/transformers ·

    Chunked prefill with prefix-cache continuation misbehaves, causing incorrect generation

    errormedium0.92 sure Would become a ticket

  40. Turbopack: next build spawns one webpack-loader process pool per loader chain × context since 16.4.0-canary.39 (hundreds of Node.js processes, OOM)

    vercel/next.js ·

    Next.js Turbopack build spawns hundreds of loader processes causing OOM

    performancehigh0.85 sure Drafted, waits for approval

  41. SynthID-Text: CPU and MPS produce different sampling tables with the same seed

    huggingface/transformers ·

    CPU and MPS backends generate different sampling tables with the same seed, indicating inconsistent watermarking behavior

    errormedium0.94 sure Would become a ticket

  42. langchain-typesafe: validate ModelRouterMiddleware classifier routes

    langchain-ai/langchain ·

    Test for ModelRouterMiddleware classifier routing fails, indicating a validation issue in langchain-typesafe

    errorlow0.92 sure Would become a ticket

  43. feat(callbacks): In-memory Zero-Trust Data Sanitization (ZTDS) callback (IETF draft-02)

    langchain-ai/langchain ·

    Request to add an in-memory Zero-Trust Data Sanitization callback handler to LangChain core

    feature requestlow0.97 sure Would become a ticket

  44. `merge_configs` drops an explicit `recursion_limit=25`

    langchain-ai/langchain ·

    merge_configs function discards the explicit recursion_limit setting

    errormedium0.94 sure Would become a ticket

  45. Turbopack dev (16.4.0): Pages Router page modules missing ("module factory is not available") after _app availability seeding (#97664)

    vercel/next.js ·

    Turbopack dev server fails to resolve Pages Router modules, causing 'module factory is not available' errors after _app seeding.

    errorhigh0.95 sure Would become a ticket

  46. Route announcer crashes with `removeChild` after browser translation

    vercel/next.js ·

    Route announcer component crashes when the page is translated in the browser

    crashmedium0.90 sure Would become a ticket

  47. Hydration mismatch false positive from Cursor's `data-cursor-ref`

    vercel/next.js ·

    Hydration mismatch warning appears due to Cursor's data-cursor-ref attribute, causing false positives in Next.js apps

    errormedium0.94 sure Would become a ticket

  48. Next 16.4: turbopackSharedRuntime inlines the bootstrap into the prerendered shell without a nonce, breaking hydration under a nonce CSP (cacheComponents)

    vercel/next.js ·

    Next.js 16.4 turbopackSharedRuntime inlines bootstrap without CSP nonce, causing hydration to break under nonce CSP

    errorhigh0.94 sure Would become a ticket

  49. langchain-typesafe: NaN/Infinity in state or criteria crash invoke with uncaught ValueError from httpx2 internals

    langchain-ai/langchain ·

    NaN/Infinity values cause uncaught ValueError in httpx2 JSON encoder, crashing TypeSafeClassifier.invoke

    crashhigh0.97 sure Would become a ticket

  50. Vectorized (batched) logits processors for assisted decoding

    huggingface/transformers ·

    Add vectorized (batched) logits processors for assisted decoding to improve efficiency

    feature requestlow0.97 sure Would become a ticket

  51. Qwen3.5-9B fine-tuning with run_clm.py on 8x B200 ran the gated DeltaNet layers on the fp32 torch chunk loop because flash-linear-attention was not installed

    huggingface/transformers ·

    User asks why run_clm.py used fp32 torch loop instead of flash-linear-attention for Qwen3.5-9B fine-tuning

    questionlow0.78 sure Drafted, waits for approval

  52. ModelCallLimitMiddleware: the limit-exceeded message can't be told apart from a model answer

    langchain-ai/langchain ·

    Limit‑exceeded messages from ModelCallLimitMiddleware are indistinguishable from normal model answers

    errormedium0.85 sure Drafted, waits for approval

  53. LukeTokenizer drops <s>/</s> on the plain-text path (no post-processor installed)

    huggingface/transformers ·

    LukeTokenizer incorrectly drops special tokens <s> and </s> when no post-processor is installed

    errorlow0.85 sure Drafted, waits for approval

  54. Improve `pretty_print`

    langchain-ai/langchain ·

    Request to enhance pretty_print to handle non-string message content like lists of content blocks more readably

    feature requestlow0.97 sure Would become a ticket

  55. `generate()` with a static cache: host-side inputs miss tensors nested in dicts, and TP undersizes the early cache

    huggingface/transformers ·

    Static cache generate() fails when inputs contain nested tensors and early cache sizing is incorrect

    errormedium0.93 sure Would become a ticket

  56. Add TokenBudgetMiddleware for token-usage budgets in agents

    langchain-ai/langchain ·

    Request to add TokenBudgetMiddleware to limit token usage in agents

    feature requestlow0.97 sure Would become a ticket

  57. QdrantVectorStore.as_retriever(search_type="mmr") inverts lambda_mult semantics

    langchain-ai/langchain ·

    User asks why QdrantVectorStore.as_retriever with search_type='mmr' inverts lambda_mult semantics

    questionlow0.78 sure Drafted, waits for approval

  58. Truncated HTML on first load leaves App Router page unhydrated with an unhandled 'Connection closed.' (React #412) rejection

    vercel/next.js ·

    Truncated HTML on first load prevents App Router page from hydrating, causing an unhandled connection closed error.

    errormedium0.90 sure Would become a ticket

  59. Nextjs test mode issue

    vercel/next.js ·

    Enabling experimental.testProxy in Next.js breaks outbound fetch calls to HTTP/2 servers

    errormedium0.87 sure Drafted, waits for approval

  60. `next/dynamic` script preloads lack `crossOrigin`, so Turbopack can't reuse them since #92070 — ~1s delay per chunk with a cross-site `assetPrefix`

    vercel/next.js ·

    next/dynamic script preloads missing crossOrigin cause ~1s delay per chunk with cross-site assetPrefix

    performancemedium0.90 sure Would become a ticket