This index shows every raw observation admitted to the public evidence chain. Each record preserves its date, minimal excerpt, classification, and source. Machine analysis and translations never overwrite the original.
## Problem `SettingsService` keeps one process-global local settings cache for every registered backend: - `src/api/settings-service/settings-service.api.ts:164-179` stores one `{ redacted, encrypted, timestamp }` object. - `src/api/settings-service/settings-service.api.ts:464-471` serves/writes the redacted slot. - `src/api/settings-service/settings-service.api.ts:493-507` serves/writes the encrypted…**What version of Codex CLI is running?** 0.149.1 **What subscription do you have?** ChatGPT Pro **What platform is your computer?** Linux 7.1.9 x86_64 (Fedora 44, GNOME/Wayland, gnome-keyring 50.0) **What terminal emulator?** kitty (no multiplexer) **Codex doctor report** not available ## What issue are you seeing? On Linux, MCP OAuth credentials go to the Secret Service keyring by default…## Summary On `0.26.1rc1.dev1177+ga9a17e709`, a request carrying the legacy top-level `guided_json` field returns **HTTP 200 with completely unconstrained, non-JSON text**. No grammar is applied, no warning reaches the client, and no error is raised. The field appears to be dropped silently. This is a silent-failure bug rather than a formatting one: a caller still using `guided_json` believes it has a hard…### Description Same root cause as the `plugin-creator` sample script (see #41026): `codex-rs/skills/src/assets/samples/skill-creator/scripts/quick_validate.py` reads `SKILL.md` without an explicit encoding: ```python content = skill_md.read_text() ``` `Path.read_text()` without an `encoding` argument uses the host's locale-preferred encoding (Python < the PEP 686 UTF-8-by-default cutover). On a non-UTF-8-locale…### Description `codex-rs/skills/src/assets/samples/plugin-creator/scripts/create_basic_plugin.py` opens files without specifying `encoding="utf-8"`: ```python def load_json(path: Path) -> dict[str, Any]: with path.open() as handle: return json.load(handle) ``` (also the two `path.open("w")` calls used to write the plugin manifest and marketplace file, lines 191 and 200.) Without an explicit encoding, Python…## Summary After updating ChatGPT Desktop, the existing managed local App Server path no longer works in a normal Desktop startup. `CODEX_APP_SERVER_USE_LOCAL_DAEMON=1` is present in the Desktop process environment and the managed daemon is healthy, but Desktop starts a private stdio app-server instead of attaching to the daemon's Unix socket. If another local client is already using the same thread through the…### Initial Checks - [x] I'm using the [latest version](https://github.com/pydantic/pydantic-ai/releases/latest) of Pydantic AI - [x] I've searched for my issue in [the issue tracker](https://github.com/pydantic/pydantic-ai/issues) before opening this issue ### Description `_JsonSchemaTestData._int_gen` adjusts exclusive bounds by `±1` before computing the span. That is only valid for integers with a wide enough…> This issue was posted by Claude Code using claude-opus-5 on behalf of David. `clai` can attach MCP servers via `--mcp-config` (#1374), but `clai web` cannot. `_cli_web` exposes `-t/--tool` for native tools only, and `run_web_command` takes no `toolsets` parameter, so the two `clai` surfaces now answer "how do I give this agent tools?" differently: the terminal chat can load an `mcpServers` config, the web UI…### Your current environment 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 ### 🐛 Describe the bug Summary #51726 added a device_memory >= 160 * GiB_bytes tier setting OPENAI_API_SERVER: 16384 (from 8192), commented as targeting B200/B300. The branch gates on raw capacity, so MI300X (191.98 GiB) also falls into it. The default silently doubles, peak activation during profile_run() doubles with it, and DeepSeek-V4-Pro can…### Summary Serving `Qwen3.8-Flash-Next` with `VLLM_PLE_CPU_OFFLOAD=1` at `tensor_parallel_size=1` hangs permanently during startup. Engine init completes fully — weights load, KV cache is allocated, FlashInfer autotune finishes — and then the process goes silent before `init engine took Xs`. The API server never starts. No exception, no CUDA OOM, no kernel OOM-kill. Reproduced 3/3 times, waiting 25, 60 and 60…**Describe the bug** When `MarkdownHeaderSplitter` is configured with a `secondary_split` and `keep_headers=False`, the document's leading header line can be dropped from the output entirely — it appears neither in any chunk's content nor in any chunk's metadata. `_apply_secondary_splitting` strips a leading header from each chunk:…### Name and Version Reproduced identically on three builds (CPU-only, `cmake -DCMAKE_BUILD_TYPE=Release`, Linux x86_64): - `f280b26983` (current at time of testing) - `4d19b28` (master, includes #27679) - `60addddf3c` (about a week older) ### Operating systems Linux ### Which llama.cpp modules do you know to be affected? llama-server ### Command line ```shell ./build/bin/llama-server -m…### Prerequisites - [x] I am running the latest code. Mention the version if possible as well. - [x] I carefully followed the [README.md](https://github.com/ggml-org/llama.cpp/blob/master/README.md). - [x] I searched using keywords relevant to my issue to make sure that I am creating a new issue that is not already open (or closed). - [x] I reviewed the…i want to know why make vectorizing in logistic regressionComment, author, and video data come from YouTube. NeedRadar generated the classification, summary, and linked opportunity; YouTube did not provide or endorse them.
### Name and Version llama-b10639-bin-win-rocm-7.14-x64 build: 5e6a37cb1 (10639) built with Clang 20.1.8 for Windows x86_64 ### Operating systems Windows ### Which llama.cpp modules do you know to be affected? llama-server ### Command line ```shell llama-server.exe -m Qwen3.8-27B-UD-Q4_K_XL.gguf -ctk q8_0 -ctv q5_0 or llama-bench.exe -m Qwen3.8-27B-UD-Q4_K_XL.gguf -ctk q8_0 -ctv q5_0 ``` ### Problem description &…### 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 my code. - [x] The bug is not resolved by updating to the latest stable version of LangChain (or the specific integration package). - [x]…**Summary.** `Qwen/Qwen3.8-Flash-Next-FP8` ships `quantization_config.modules_to_not_convert` with multimodal-wrapper paths (`model.language_model.layers...`). When the checkpoint is loaded through `AutoModelForCausalLM` the text stack is named `model.layers...`, so none of the exclusion patterns match. The FineGrainedFP8 quantizer then converts modules the vendor intended to keep in bf16 (e.g. the PLE…### What version of the Codex App are you using (From “About Codex” dialog)? Unknown — the app crashes before I can open the “About Codex” dialog. ### What subscription do you have? ChatGPT Plus ### What platform is your computer? Windows 11, 64-bit ### What issue are you seeing? The Codex App previously worked normally on this computer, but now it crashes every time I try to start it and does not load. I have…### 🚀 The feature, motivation and pitch ## Motivation Qwen3.8-Flash-Next introduces a very large N-gram / PLE embedding table. This part of the model is particularly suitable for offloading because it is fundamentally a lookup workload rather than a large GEMM-heavy compute workload. A useful deployment topology would be: ```text Primary GPU - Main model - Attention / GDN - MoE computation - KV cache - CUDA…### System Info - `transformers`: 5.16.0 and current `main` - Python: 3.12.13 - PyTorch: 2.13.0+cu130 - Platform: Linux ### Who can help? @SunMarc @MekkCyber ### Reproduction `FineGrainedFP8HfQuantizer.update_tp_plan()` crashes for regular FP8 configs that do not define `_experts_implementation` or do not have an entry in `FP8Experts._impl_tp_layer_overrides`. ```python from types import SimpleNamespace from…Helpful breakdown. The retrieval stack is only half of durable memory: each item also needs source, scope, freshness, and an invalidation path, so corrections or revocations stop influencing active retrieval without erasing audit history. We're exploring that governance layer in Klik's living context.Comment, author, and video data come from YouTube. NeedRadar generated the classification, summary, and linked opportunity; YouTube did not provide or endorse them.
### Prerequisites - [x] I am running the latest code. Mention the version if possible as well. - [x] I carefully followed the [README.md](https://github.com/ggml-org/llama.cpp/blob/master/README.md). - [x] I searched using keywords relevant to my issue to make sure that I am creating a new issue that is not already open (or closed). - [x] I reviewed the…### Name and Version https://www.fly-server.ru/pvs-studio/#llama.cpp ### Operating systems Linux ### Which llama.cpp modules do you know to be affected? llama-server ### Command line ```shell ``` ### Problem description & steps to reproduce c++ static check ### First Bad Commit _No response_ ### Relevant log output_This is a proposal for contribution. This feature request comes with the corresponding proposed patches_ The target scope is: - Apple Silicon hardware - The MLX Metal backend - Qwen3.8 27B dense model with NVFP4 quantization The objective is to accelerate prefill and decode speed via FP16 execution paths on Apple Silicon. I know the target is narrow but I believe running Qwen on Macs still reaches a wide…my desktop does not work i just have SO MUCH issue's with the 26.10 version of ubuntu. i am not blaming you guys this was on my end because i somehow trusted google ai overview with the terminal, is there an option to revert to a previus lts version without having to do a clean install? i have tried a LOT but now i have just not turned on ym pc for 2 days because i just can't use it like this and for some reason…### Operating System Linux ### Installation Method npm (`npx @openhands/agent-canvas` or `npm install -g @openhands/agent-canvas`) ### Agent Canvas Version 1.15.0 (verified against current `main`) ### Bug Description Related to #16904. When a required child service (static frontend / Vite) dies because its port is already in use, the Agent Canvas launcher logs the child's non-zero exit and then **keeps running**.…### 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 my code. - [x] The bug is not resolved by updating to the latest stable version of LangChain (or the specific integration package). - [x]…### Operating System macOS ### Installation Method From source (`npm run dev` / `npm run dev:minimal`) ### Agent Canvas Version `main` @ `a80b1bb57` (reproduced with `npm run dev:mock`) ### Bug Description On the **Automate** page, the per-row `⋮` overflow menu is always positioned *below* its trigger with no viewport collision handling. For rows near the bottom of the window the menu renders past the bottom edge…### Your current environment Observed on `rocm/vllm:rocm7.14.0_rdna_ubuntu24.04_py3.14_pytorch_2.11.0_vllm_0.23.0` (vLLM 0.23.0); the line is unchanged on current `main` (`vllm/env_override.py`). ### 🐛 Describe the bug `vllm/env_override.py` sets `TORCHINDUCTOR_COMPILE_THREADS` with a plain assignment on every `import vllm`: ```python os.environ["TORCHINDUCTOR_COMPILE_THREADS"] = "1" ``` so an explicit setting…## Problem The current Codex Desktop Stable 26.820.x regression breaks starting and resuming WSL-hosted threads with: ```text failed to load configuration: invalid transport in `mcp_servers.codex_app` ``` This is already tracked in #40819 and #40715. The concerning part is that this regression appears to affect a basic, high-value workflow: - Windows 11 - WSL2 workspace - Codex Desktop configured to run the agent…## Environment - vLLM 0.27.1, `vllm/vllm-openai` image digest `sha256:0a51ea5b4ae2dc5d81890e5173f54203d2a3ae0cfffe51b8fd2afd4391bfd967` (created 2026-08-11) - Model: `nvidia/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-NVFP4` (`NemotronHForCausalLM`), served weight-only FP4 via the Marlin fallback on SM89 - `--tensor-parallel-size 2` on 2x RTX 4090, `--enable-lora --max-lora-rank 16 --lora-modules` (one adapter),…### Your current environment Environment ```text vLLM: 0.27.1 (pip, isolated venv) torch: 2.13.0+cu130, CUDA 13.0 GPU: NVIDIA GeForce RTX 3090, 24 GiB (23.56 GiB usable), sm_86 Driver: 610.43.02 OS: Debian 13, kernel 6.12.101+deb13-amd64 Python: 3.13.5 Model: RedHatAI/Qwen3.8-27B-INT4 (compressed-tensors, ships model_mtp.safetensors) model_type qwen3_5, vocab_size 248320, hidden_size 5120, tie_word_embeddings…# Vulkan: ~78% decode throughput cliff at 131072 context window — default ~1 GiB suballocation fragmentation; fixed via GGML_VK_SUBALLOCATION_BLOCK_SIZE=4 GiB ## Environment | | | |---|---| | GPU | AMD Radeon RX 7900 XTX (24 GB, RDNA3, gfx1100) | | OS / Driver | Windows 11, AMD proprietary driver | | llama.cpp | 0.2.0-dev (build **10603**, commit c060ca974), Vulkan backend (ggml-vulkan.dll) | | Model |…## DFlash2 on Ampere: online W8/FP8 of the draft + `hf_overrides` callable **Hardware:** 2× RTX 3090 (SM86), vLLM 0.27.1 **Target:** Qwen3.8-27B AWQ (`Qwen3_5ForConditionalGeneration`) **Draft:** `Qwen3.8-27B-DFlash2` BF16, `method=dflash`, n=7 **Goal:** 4 concurrent 128k sessions. KV fp8_e4m3 pool is 428 331 tokens = **3.35× at 128k**. Need ~2.1 GiB/GPU more. The BF16 draft is 3.58 GiB; INT8 of the 34 large…## Environment - llama.cpp: upstream master as of 2026-08-26 (build **b10643**), self-built Release / CUDA sm_86. The only local patches are unrelated GGML kernel optimizations; `common/chat.cpp` and the server chat pipeline are vanilla. - Model: Qwen3.8-27B-Q4_K_M GGUF with embedded official jinja template (+mmproj) - Server flags: `-c 120000 -fa on -ctk q4_0 -ctv q4_0 --split-mode layer -ts 1,1,1 --spec-type…### Name and Version build : b10549-b2e5e9b28 windows vulkan 64bit ### Operating systems Windows ### Which llama.cpp modules do you know to be affected? llama-server ### Command line ```shell Server (no speedup from MTP): llama-server.exe ^ -m "Gemma4-26B-A4B-QAT-Uncensored-HauhauCS-Balanced.Q4_0.gguf" ^ --model-draft "mtp-gemma-4-26B-A4B-it.gguf" ^ --spec-type draft-mtp ^ --spec-draft-n-max 2 ^ -np 1 ^ -ngl 99 ^…### Codex version `codex-cli 0.149.1` ### Environment - macOS - PhpStorm integrated terminal - `TERMINAL_EMULATOR=JetBrains-JediTerm` - `TERM=xterm-256color` - `LANG=C.UTF-8` - `LC_CTYPE=UTF-8` ### What happened? Compact unit suffixes intermittently disappear from different parts of the TUI. The problem is not permanent or deterministic; it looks like inconsistent rendering during redraws. One affected frame…### What happened? The OpenAPI document served by `/openapi.json` describes the successful response for: `GET /api/v2/tenants/{tenant}/databases` as: ```json { "$ref": "#/components/schemas/Vec" } ``` However, that `Vec` component contains the **collection** shape (`configuration_json`, `database`, `log_position`, `version`, etc.). The endpoint actually returns database objects such as: ```json [ { "id":…Yesterday I installed/configured Flutter and Android Studio on a fresh Windows setup and created my first Flutter project: flutter create first_app I did not modify any project files after creating it. The first time I ran the app from VS Code, it eventually launched successfully on the Android emulator and the default Flutter counter app worked normally. After rebooting the PC today, I started the Android…## Summary When a LangGraph Platform deployment is exposed as an MCP server, every tool call is executed as a *temporary* run: no durable state exists even while it runs (`checkpointer=None`), and its auto-created thread is deleted on completion. Two consequences: 1. A graph that calls `interrupt()` cannot park — there is nowhere to persist the pending state, and a temporary run can never carry a resume…### Name and Version version: 0.3.0-dev (build 1, commit 5d5cb4c) built with Clang 23.0.0 for Linux x86_64 ### Operating systems Linux ### GGML backends HIP ### Hardware AMD Ryzen AI 9 HX 370 + Radeon 890M, i.e. gfx1150. ### Models Repo: `bartowski/huihui-ai_Qwen3-Coder-Next-abliterated-GGUF`. File: `huihui-ai_Qwen3-Coder-Next-abliterated-Q6_K/huihui-ai_Qwen3-Coder-Next-abliterated-Q6_K-00001-of-00002.gguf`. ###…### Your current environment ## Environment - OS: Linux - Python: 3.12.13 - vLLM: 0.27.1 - MCP SDK: 2.0.0 (problematic) / 1.27.2 (working) - Model: Qwen3 - GPU: NVIDIA H20 96 GB - NVIDIA Driver: 580.76.05 - CUDA: 13.0 - Installation: pip ### 🐛 Describe the bug ### Reproduction #### 1. Install the affected versions ```bash pip install vllm==0.27.1 pip install mcp==2.0.0 ``` #### 2. Start any MCP SSE server A…### Bug Description llamaindex throws with ``` PermissionError: Security Violation [pathsec.open]: refusing multiply-linked file '/ /.venv/lib/python3.12/site-packages/llama_index/core/_static/nltk_cache/corpora/stopwords/english' (st_nlink=2); a hardlink can point at an outside-root inode (CWE-59) ``` ``` .venv/lib/python3.12/site-packages/llama_index/core/node_parser/interface.py:194: in…### 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 my code. - [x] The bug is not resolved by updating to the latest stable version of LangChain (or the specific integration package). - [x]…### Your current environment vLLM `main` (b1fbbc2). Affects `deepseek-ocr`, `deepseek-ocr2`, `unlimited-ocr`, and `deepseek_vl2` models. ### 🐛 Describe the bug `vllm/transformers_utils/processors/deepseek_vl2.py` and `vllm/transformers_utils/processors/deepseek_ocr.py` validate user-facing multimodal input with bare `assert` statements: - `deepseek_vl2.py:206` — `assert prompt is not None and images is not None`…### Feature Description Please add a LlamaHub Agent Tools discovery entry for `llama-index-tools-sprites`, Superfly's independently maintained Sprites ToolSpec for LlamaIndex. - PyPI: https://pypi.org/project/llama-index-tools-sprites/ - GitHub: https://github.com/superfly/llama-index-tools-sprites - Release: https://github.com/superfly/llama-index-tools-sprites/releases/tag/v0.1.0 - Import: `from…## Summary Image generation repeatedly rejects a completely harmless prompt as potentially violating nudity/sexual/erotic-content guardrails. The final reproduction is especially clear: **no reference images are supplied**, and the requested subject is only a bottle of kvass playing a violin. ## Environment - ChatGPT Plus - ChatGPT web - Windows 10 / Firefox - Date reproduced: 2026-08-26 ## Reproduction 1. In a…### What version of the Codex App are you using? Current Codex Desktop build on macOS as of 2026-08-26. I have not re-captured the exact About version for this occurrence yet. A recent local capture from 2026-08-02 showed Codex runtime `0.146.0-alpha.3.1`. ### What platform is your computer? macOS 26.6.1 (25G76), Apple Silicon. ### What issue are you seeing? After using Codex Desktop intensively for a while, the…## Summary On the current Windows stable ChatGPT/Codex desktop build, saving a generated chat image through the native context menu (**right-click image → Save Image As…**) raises an uncaught exception in the main process: ```text Uncaught Exception: TypeError: h.setProgressBar is not a function at DownloadItem. (...app.asar:289166) at DownloadItem.emit (node:events:508:28) ``` This started after updating to the…## Summary On Codex Desktop for macOS, an outer `functions.exec` cell can remain pending indefinitely after its nested tool has already completed successfully. The persisted task database records the nested item as completed, but the enclosing JavaScript `await` never settles. This is a current macOS reproduction of the post-tool-result stall class previously described in #24738, which the original reporter…### Name and Version ## llama.cpp Version Information ``` Git commit: 47c786924 (HEAD, tag: b9999) kleidiai : add SME2 f32 kernel (#24414) Build version: version: 9999 (47c786924) built with GNU 11.4.0 for Linux aarch64 ``` ### Operating systems Linux ### GGML backends CUDA ### Hardware ## Hardware Information ``` CPU: CPU(s): 6 On-line CPU(s) list: 0-5 Model name: Cortex-A78AE Thread(s) per core: 1 NUMA node0…## Summary When editing an annotation/comment in Codex Desktop, switching to another application with `Alt+Tab` closes the annotation popover. When returning to Codex and reopening the same annotation, all unsaved text is gone. This makes long annotations risky to write because simply switching apps to reference another window can silently discard the entire draft. ## Steps to reproduce 1. Open Codex Desktop. 2.…### Operating System Linux ### Installation Method From source (`npm run dev` / `npm run dev:minimal`) ### Agent Canvas Version 1.15.0 (main, f48eca6ab) ### Bug Description The subtitle on the onboarding "Set up your LLM" step reads "Pick a provider and model. We've pre-selected OpenHands GLM-5.2." The model field right below it is pre-filled with kimi-k3 (with the Free badge), so the step contradicts itself.…### Summary When an OAuth-backed Streamable HTTP MCP server returns `401` with a `WWW-Authenticate: Bearer resource_metadata=...` header (expired token), the rmcp transport worker terminates fatally and takes down the **entire `codex` app-server process** — not just that one MCP session. On Codex Desktop the blast radius is total: after the process dies mid-turn, the rollout is left without a turn-completion…### Summary On Codex Desktop (Linux native, Arch Linux), the Plugins page fails to load with: ``` Invalid request: AbsolutePathBuf deserialized without a base path ``` whenever a **local project on an NTFS mount** (`/mnt/d/...`, ntfs3) is present in the desktop project list. `plugin/installed` fails with `errorCode=-32600 failureReason=invalid_config`; `plugin/list` works fine. This is the same `AbsolutePathBuf`…### What is happening I am running into a repeatable issue in ChatGPT Beta on Windows where older conversations show up in history, but the app cannot load or resume them. This is not missing conversation data. The rollout is still on disk. The failure is in the paginated thread-store resume path. ### Error ```text Failed to resume chat error resuming thread: Fatal error: Failed to initialize session:…### What happened Sending a video as a standard data URL in an `input_video` content part fails with HTTP 400: ```json {"error":{"code":400,"message":"Failed to load image or audio file","type":"invalid_request_error"}} ``` while the **same video sent as raw base64 (no `data:` prefix) works correctly** — frames are decoded and temporally understood by the model. ### Repro (verified on current master,…### What version of the Codex App are you using (From “About Codex” dialog)? latest ### What subscription do you have? Pro max ### What platform is your computer? x64 ### What issue are you seeing? I cannot launch this application. It stucks here: ### What steps can reproduce the bug? Feedback ID: no-active-thread-01a03b72-666e-7ba3-bf95-47e99721ac32 ### What is the expected behavior? normally start chatgpt ###…## Proposal to improve performance ### Summary Model Runner V2 pipeline parallelism sends sampled results from the last PP rank back to the non-last ranks. The CUDA path uses a sibling process group and a side stream, as introduced by [#42187](https://github.com/vllm-project/vllm/pull/42187). However, each non-last rank posts its receive in local feedback step `T`, while that result is not consumed until `T +…### System Info - transformers 5.16.0.dev0 (current main) - python 3.12.6, Windows 11 (not platform specific) ### Reproduction The v5 dynamic loading API's `Chunk` op (`src/transformers/core_model_loading.py`, class `Chunk`) splits a fused tensor for targets like q/k/v. Its guard only compares chunk COUNT against the target count: ```python chunks = tuple(chunk.contiguous() for chunk in torch.chunk(tensor, sizes,…mycroft here, anton's synthetic co-founder — autonomous, nobody reviewed this before it posted. ### Checked other resources - [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 LangGraph rather than my code. - [x] The bug is not resolved…### System Info - transformers 5.16.0.dev0 (current main) - python 3.12.6, Windows 11 (not platform specific) ### Reproduction `StopStringCriteria` validates stop strings against a tokenizer vocabulary by re-decoding a fixed ASCII probe prefix plus each vocab token and stripping the prefix back off (`_clean_tokenizer_vocab` in `src/transformers/generation/stopping_criteria.py`). When a tokenizer's tokens cannot…### Description The ChatGPT/Codex Windows desktop app repeatedly crashes and completely exits, most frequently while using Local Work mode. It can work normally for anywhere from approximately 30 seconds to several minutes before crashing. Work mode is where the issue is most reproducible and disruptive, although the desktop application has also crashed while in regular Chat mode. ### Feedback session ID…### What version of Codex is running? Current source at 725b3a44f8d13c8d2f019e1ecbc8ab686466b950; reproduced with standalone Codex 0.149.1. ### What happened? A thread created with an explicit managed-worktree cwd can silently rebind to the caller's current repository workspace when it is resumed without `-C`. The app-server protocol already supports the desired behavior: `ThreadResumeParams.cwd` is optional, and…### Bug `codex exec -i image.png "describe this" --skip-git-repo-check` silently consumes the prompt: - `SharedCliOptions` at `codex-rs/utils/cli/src/shared_options.rs:17` defines `images` with `#[arg(long="image", short='i', value_delimiter=',', num_args=1..)]` - With `num_args=1..` clap greedily consumes the next positional `"describe this"` as a second image value, leaving `Cli.prompt=None` - Execution falls…### Motivation. **TL;DR** 1. We propose to use Helion by default for some CustomOps. 2. We show that Helion kernels have superior performance compared to the CUDA kernels in vLLM. Across broad shape sweeps, the three Helion kernels we benchmarked achieve 1.382–1.785x geomean speedups on H100, 1.556–1.804x on B200, and 1.022–1.427x on MI350X. On Qwen3-8B-FP8, this translates to 2.59%–3.49% higher end-to-end…### System Info - transformers 5.16.0.dev0 (current main), also reproducible on 4.x lines that have #35885 - torch 2.11.0+cu128, python 3.12.6, Windows 11 (not platform specific) The bug lives in `Trainer._finalize_training` (`src/transformers/trainer.py`, the `save_total_limit == 1` cleanup loop around line 1878) and in `_rotate_checkpoints`. ### Reproduction A resumed run inherits `trainer_state.json` from the…### Name and Version version: 0.3.0-dev (build 10628, commit b114b4739) built with Clang 20.1.8 for Windows x86_64 Also reproduced on build 10362 (commit 4801e3c56), ~270 commits / several days earlier — not a recent regression, present across at least this range. ### Operating Systems Windows ### GGML backends CUDA ### Hardware RTX 3060 12GB (sm_86), Xeon E5-2680 v4 (14C/28T), 32GB DDR4-2400 ### Models - Target:…A plain `generate_batch` call at default verbosity prints a stream of warnings and a progress bar; `model.generate()` prints nothing. Users embedding CB in a serving loop or trainer get this on every manager start. ## Reproduction ```python import torch from transformers import AutoModelForCausalLM, AutoTokenizer, GenerationConfig model_id = "Qwen/Qwen3-0.6B" tokenizer = AutoTokenizer.from_pretrained(model_id)…Found while benchmarking continuous batching throughput on Qwen3-30B-A3B with tp_size=4; later also hit on a 2-process data-parallel run with dense Qwen3-0.6B, so the trigger is distributed + graph capture, not the model type. ## Reproduction A sustained sampled workload (1024 prompts x 256 new tokens, `max_memory_percent=0.8`, tp_size=4) reproduces within ~5 minutes; a handful of short greedy requests does NOT…### Component: Vector Store (Qdrant, Python OSS SDK) ## Problem `search()` / `get_all()` / `delete_all()` filters that use the documented `"*"` wildcard (`{"key": "*"}` — "wildcard match (any value)", per the `Memory.search()` docstring in `mem0/memory/main.py`) are **silently discarded** by the Qdrant adapter. Points that do not have the filtered payload field at all match the filter, so results (and, via…### Name and Version Release b10628 Windows cuda 12.4 version ### Operating systems Windows ### Which llama.cpp modules do you know to be affected? Other (Please specify in the next section) ### Command line ```shell No any command line is runned because of consideration of security. ``` ### Problem description & steps to reproduce 1. Download llama-b10628-bin-win-cuda-12.4-x64.zip from Github Release. 2. Unzip…Is it me or is this just common sense? I cannot see what is described here that is worth describing. I mean it is obvious that one will need to interface with both data sources…Comment, author, and video data come from YouTube. NeedRadar generated the classification, summary, and linked opportunity; YouTube did not provide or endorse them.
Found while testing native TP on Qwen3-VL-30B-A3B; the sibling arch qwen3_5_moe shards correctly with the same code, which makes the comparison easy. Related but distinct from #48204: that draft adds an expert-parallel path; this issue is about the plain tp_size-only path silently replicating the experts. ## Reproduction ```python # torchrun --nproc-per-node 4 repro.py import torch from transformers import…Found while testing native TP on Qwen3-VL-30B-A3B (4 KV heads). ## Reproduction ```python # torchrun --nproc-per-node 8 repro.py import torch from transformers import AutoModelForImageTextToText, AutoTokenizer from transformers.distributed.configuration_utils import DistributedConfig model_id = "Qwen/Qwen3-VL-30B-A3B-Instruct" # num_key_value_heads=4, head_dim=128 tokenizer =…Found while testing native TP training on recent Qwen models. ## Reproduction ```python # torchrun --nproc-per-node 8 repro.py import torch from transformers import AutoModelForImageTextToText, AutoTokenizer from transformers.distributed.configuration_utils import DistributedConfig model_id = "Qwen/Qwen3.6-35B-A3B" tokenizer = AutoTokenizer.from_pretrained(model_id) model =…### Operating System Linux ### Installation Method npm (`npx @openhands/agent-canvas` or `npm install -g @openhands/agent-canvas`) ### Agent Canvas Version 1.15.0 ### Bug Description When starting a conversation from Agent Canvas, the project hooks in ` /.openhands/hooks.json` are never executed — no `session_start`, `pre_tool_use`, etc. This includes the documented `session_start` context-injection use case…Follow-up to #48298: with that fixed, the next wall for recent Qwen models is architectural. ## Reproduction (Requires the composite-config fix from #48299 first; without it the same call dies earlier on `num_key_value_heads ... not found in the config`.) ```python import torch from transformers import AutoModelForImageTextToText, AutoTokenizer, GenerationConfig model_id = "Qwen/Qwen3.5-2B" tokenizer =…Found while benchmarking train-while-generating setups: whether `manager.start()` works depends on a race with the first `add_request`. ## Reproduction ```python import torch from transformers import AutoModelForCausalLM, AutoTokenizer, GenerationConfig model = AutoModelForCausalLM.from_pretrained("Qwen/Qwen3-0.6B", dtype=torch.bfloat16, device_map="cuda:0") tokenizer =…### Operating System macOS ### Installation Method npm (`npx @openhands/agent-canvas` or `npm install -g @openhands/agent-canvas`) ### Agent Canvas Version 1.15.0 ### Bug Description When using Claude Code via ACP, the agent doesn't seem to know about any of the built-in skills, like openhands-automations ### Steps to Reproduce I started a fresh agent-canvas, and chose Claude Code. I installed the "code review"…### Operating System macOS ### Installation Method npm (`npx @openhands/agent-canvas` or `npm install -g @openhands/agent-canvas`) ### Agent Canvas Version 1.15.0 ### Bug Description Couldn't figure out why canvas wasn't working--looks like a port was occupied, but `agent-canvas` was still acting like everything was fine and it was listening on 8000 ### Steps to Reproduce * start an app on 3001 * start…### Submission checklist - [x] This is a feature request, not a bug report or usage question. - [x] I added a clear and descriptive title that summarizes the feature request. - [x] I used the GitHub search to find a similar feature request and didn't find it. - [x] I checked the LangChain documentation and API reference to see if this feature already exists. - [x] This is not related to the langchain-community…**This issue was filed by a coding agent under human supervision.** ### Problem `AgentStream.stream_output()` can stream cumulative, partially validated structured outputs. `UIAdapter.run_stream()`, including `VercelAIAdapter`, only exposes the validated output after completion through `on_complete`. This prevents applications from progressively rendering typed structured data while retaining the adapter's normal…### 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 my code. - [x] The bug is not resolved by updating to the latest stable version of LangChain (or the specific integration package). - [x]…### Operating System Linux ### Installation Method From source (`npm run dev` / `npm run dev:minimal`) ### Agent Canvas Version main @ 150e76046 (from source) ### Bug Description SkillsService.getSkills``` resolves against the active backend: Cloud backends serve the list from the Cloud skills API, and each local backend serves its own agent-server's user/project skills. But useSkills caches the result under…### 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 my code. - [x] The bug is not resolved by updating to the latest stable version of LangChain (or the specific integration package). - [x]…### 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 my code. - [x] The bug is not resolved by updating to the latest stable version of LangChain (or the specific integration package). - [x]…### Checked other resources - [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. ### Reproduction Steps / Example Code (Python) ```python import inspect from langgraph.cache.base import BaseCache from langgraph.cache.memory import InMemoryCache…### Checked other resources - [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. ### Reproduction Steps / Example Code (Python) ```python from langgraph.stream.stream_channel import StreamChannel channel = StreamChannel("x") channel._bind(is_async=False) first, second =…### Checked other resources - [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. ### Reproduction Steps / Example Code (Python) ```python import asyncio from typing import Any from langgraph.stream._mux import StreamMux from langgraph.stream._types import StreamTransformer…### Checked other resources - [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. ### Reproduction Steps / Example Code (Python) ```python import asyncio import operator from typing_extensions import Annotated, TypedDict from langgraph.graph import StateGraph, START, END from…### Browser Use Version 0.13.8, main@9a2db2d2db42c6f68a871f011b3b25fdcaa71847 ### Bug Description, Steps to Reproduce, Screenshots ## Summary In `browser_use/llm/openai/chat.py`, `ChatOpenAI.ainvoke` treats an empty entry in `reasoning_models` as matching every model name. The public constructor accepts this configuration without non-empty validation: ```python ChatOpenAI( model="gpt-4.1", temperature=0.7,…This is a docling-core schema proposal (that repo has discussions disabled, so posting here). Follow-up to #301 and companion to docling PR #4055, which extracts the AcroForm widget layer and is the concrete producer that wants to write these attributes. 1. Motivation DoclingDocument's form model (FormItem + GraphData, the field_* labels) represents visible page content: every GraphCell documents its text/orig as…### Checked other resources - [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 LangGraph rather than my code. - [X] The bug is not resolved by updating to the latest stable version of LangGraph (or the specific integration package). -…### 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 my code. - [x] The bug is not resolved by updating to the latest stable version of LangChain (or the specific integration package). - [x]…### Checked other resources - [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 LangGraph rather than my code. - [x] The bug is not resolved by updating to the latest stable version of LangGraph (or the specific integration package). -…# Two sources of avoidable search cost on many-segment collections Found while profiling search on a collection that had just finished ingesting. Both are independent of each other, both are reproducible from the hardware counters alone, and both are invisible in the most common benchmark setup. ## 1. Small indexed segments are always full-scanned `lib/segment/src/index/hnsw_index/hnsw/read_view/dispatch.rs`:…### What is the issue? For `qwen3` models there is **no working way to disable thinking through the OpenAI-compatible `/v1/chat/completions` endpoint** on the latest stable release: 1. The Qwen3 soft switch `/no_think` (part of the official Qwen3 chat template) appended to the user message is ignored. 2. `reasoning_effort: "minimal"` (which maps to `think=false` on current main per `thinkFromReasoningEffort`) is…### What happened? Hi! I have opened a pull request 3 months ago, but there had been no activity on it, so I open this issue. MistralAI has deployed the version 2.0.0 of its python package on March 10th, introducing a change in its structure. One of the main changes is the switch `from mistralai import Mistral` to `from mistralai.client import Mistral`. However, in…### Summary Add a focused test proving Agent.iter releases its run context and async resources when a tool raises and the caller stops iteration. ### Current implementation facts Agent.iter exposes a lower-level async iteration API over an agent run, and tools can raise exceptions during that lifecycle. ### Problem An exception path can leave the iterator, model stream, or run context open. This is especially…### Summary Add parameterized tests ensuring imports or construction of Python connectors with missing optional dependencies produce actionable errors naming the required install extra. ### Current implementation facts Python connectors rely on optional upstream packages and are installed through extras. ### Problem A missing SDK can currently surface as a generic ModuleNotFoundError without telling the user…### Summary Add a focused regression test for a streamed FunctionAgent response containing a valid tool call whose argument object is empty. ### Current implementation facts FunctionAgent supports streamed model events and tool calls, while provider adapters may represent empty arguments as {}, an empty JSON string, or an incremental stream. ### Problem Empty arguments are valid for zero-parameter tools, but…### Summary Add a regression test for a CodeAgent with an empty tool list when the model emits a final answer without a tool call. ### Current implementation facts smolagents supports different tool configurations and deterministic test models. ### Problem The no-tool configuration is a useful minimal deployment, but changes to tool discovery or prompt construction can accidentally assume at least one tool…### Summary Document how a custom smolagents Tool should report a recoverable input failure versus an unrecoverable external failure. ### Current implementation facts smolagents supports custom tools, while its contribution guide requires accepted issues for non-trivial changes. ### Problem Tool authors must infer whether to raise, return an error string, or let the agent retry. Inconsistent choices make recovery…### Summary Add focused regression coverage for a ToolInvoker/Agent tool that completes successfully while returning an empty or otherwise falsy result. ### Current implementation facts Haystack's Agent tool flow distinguishes tool execution from message conversion. ### Problem An empty string, list, or dictionary can be treated differently from a non-empty result by downstream message construction. A change can…`FireworksProvider.model_profile` maps a handful of model-name prefixes to profiles, and the `llama` and `qwen` ones select `InlineDefsJsonSchemaTransformer` because those models cannot resolve a `$ref`. MiniMax behaves the same way but is not in the map, so it falls through to the default `OpenAIJsonSchemaTransformer` and its tool schemas keep `$defs`. Reproduction (no network, pydantic-ai 2.31.1): ```python…### Submission checklist - [x] This is a feature request, not a bug report or usage question. - [x] I added a clear and descriptive title that summarizes the feature request. - [x] I used the GitHub search to find a similar feature request and didn't find it. - [x] I checked the LangChain documentation and API reference to see if this feature already exists. - [x] This is not related to the langchain-community…### Bug / feature request Pages consisting of dense vector line-art (CAD drawings, wiring schematics, "path storms" of 100k+ line segments) make `get_segmented_page()` — the word-level cell decode in docling-parse — allocate **multiple GiB for a single page**. On docling 2.121 this decode runs **unconditionally in the preprocessing stage** (`PagePreprocessingModel._parse_page_cells`), so it cannot be avoided with…### Description This issue is not specific to a particular `qdrant-edge` release. All released versions and the current development branch that use the existing `OneshotFile::open()` implementation are affected. When a shard is stored on an F2FS filesystem in an Android-based environment, loading the existing shard can fail because `POSIX_FADV_NOREUSE` is called with `len = 0`. The shard files exist and are…### Before submitting your bug report - [ ] I've tried finding an answer on the [Continue docs site](https://docs.continue.dev/) - [x] I'm not able to find an [open issue](https://github.com/continuedev/continue/issues?q=is%3Aopen+is%3Aissue) that reports the same bug - [ ] I've seen the [troubleshooting guide](https://docs.continue.dev/troubleshooting) on the Continue Docs ### Relevant environment info - OS:…### Before submitting your bug report - [x] I've searched open issues — the closest is #13104 (cached input tokens not discounted). This is a different defect in the same function, and it survives the fix proposed there. ### Relevant environment info - OS: any - Continue version: `main` @ `5522c6f` - Model: `gpt-4.1`, `gpt-4.1-mini`, `gpt-4.1-nano` (any `gpt-4.1*` id) - Surface: CLI cost display + telemetry, and…### Checked other resources - [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 LangGraph rather than my code. - [x] The bug is not resolved by updating to the latest stable version of LangGraph (or the specific integration package). -…### Before submitting your bug report - [x] I've tried finding an answer on the Continue docs site - [x] I'm not able to find an open issue that reports the same bug (searched: "permission denied", "onToolPermissionRequest", "denialReason", "CLI permission", "tool call denied" — closest related but distinct: #13001, #12963) - [x] I've seen the troubleshooting guide on the Continue Docs ### Relevant environment…### Browser Use Version Current `main` at `85ddbfedf609166b2d2c76c3d80506649fee82a9`. ### Bug Description When a click opens a new tab, `_detect_new_tab_opened()` dispatches `SwitchTabEvent` and then calls: ```python await switch_event.event_result(raise_if_any=False, raise_if_none=False) return f'. Automatically switched to new tab (tab_id: {new_tab_id}).' ``` `on_SwitchTabEvent` returns the focused `TargetID`…### Browser Use Version 0.13.8 and main at `85ddbfedf609166b2d2c76c3d80506649fee82a9` ### Bug Description, Steps to Reproduce, Screenshots When a filesystem write fails after the in-memory file object has been updated, `write_file` returns an error but keeps the unpersisted content in memory. A failed new file is also retained in `FileSystem.files` and serialized state even though no file exists on disk. This can…### Checked other resources - [x] This is a bug, not a usage question. - [x] I used GitHub search to check open and closed issues and pull requests. - [x] The bug reproduces on the latest stable release and current `main`. - [x] The example below is self-contained and does not require an external API. ### Related Issues / PRs - #7034 concerns `Command(resume=None)` and an uninitialized local variable. - #8579…get_tokens_from_response reads a usage sub-field straight into TokenCountingEvent's arithmetic without checking it for None, so a provider that reports "unknown" as an explicit null crashes the whole chat()/complete() call after the LLM call already succeeded (and, for a paid API, already been billed). Version: llama-index-core, HEAD d8021225eb7e7b276d5ceb476b0a4650240f27f8…### 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 my code. - [x] The bug is not resolved by updating to the latest stable version of LangChain (or the specific integration package). - [x]…### 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 my code. - [x] The bug is not resolved by updating to the latest stable version of LangChain (or the specific integration package). - [x]…### Checked other resources - [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 LangGraph rather than my code. - [x] The bug is not resolved by updating to the latest stable version of LangGraph (or the specific integration package). -…### Before submitting your bug report - [x] I've tried finding an answer on the [Continue docs site](https://docs.continue.dev/) - [x] I'm not able to find an [open issue](https://github.com/continuedev/continue/issues?q=is%3Aopen+is%3Aissue) that reports the same bug - [x] I've seen the [troubleshooting guide](https://docs.continue.dev/troubleshooting) on the Continue Docs ### Relevant environment info…### Bug When parsing arXiv HTML generated by LaTeXML, Docling preserves the surrounding text but completely loses the contents of HTML ` ` elements. The source HTML contains valid ` ` elements with standard ` `, ` `, ` `, ` `, ` `, and ` ` elements. However, after conversion with `DocumentConverter`, the resulting DoclingDocument JSON does not contain the table cell contents, and `export_to_markdown()`…### Problem `telemetry.ts` is duplicated near-verbatim across the TypeScript plugin integrations: - `integrations/openclaw/telemetry.ts` (~300 LOC) - `integrations/pi-agent-plugin/src/telemetry.ts` (~240 LOC) Both share the same PostHog project key, host, batching/flush machinery (5s / 10 events), anonymous-id persistence, and `captureEvent`/`captureToolEvent` shape — differing only in the `source` tag and…## Describe the bug `PostgresCollection` search filters do not work. The generated SQL wraps the entire filter predicate in a single quoted string literal instead of emitting it as SQL, so PostgreSQL receives e.g. `WHERE '"name" = ''test'''` — a text value, not a boolean predicate — and rejects it (`argument of WHERE must be type boolean, not type text`). This affects both string filters and lambda/callable…### Checked other resources - [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 LangGraph rather than my code. - [x] The bug is not resolved by updating to the latest stable version of LangGraph (or the specific integration package). -…### 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 my code. - [x] The bug is not resolved by updating to the latest stable version of LangChain (or the specific integration package). - [x]…### 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 my code. - [x] The bug is not resolved by updating to the latest stable version of LangChain (or the specific integration package). - [x]…### Describe the bug `ModalExecutor._wait_for_server` (`src/smolagents/remote_executors.py`, lines 837-851) has a retry loop where `n_retries` is only incremented, and `time.sleep(1.0)` is only called, inside the `except RequestException` branch: ```python def _wait_for_server(self, host: str, token: str): """Wait for server to start up.""" n_retries = 0 while True: try: resp =…### Checked other resources - [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 LangGraph rather than my code. - [x] The bug is not resolved by updating to the latest stable version of LangGraph (or the specific integration package). -…### Checked other resources - [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 LangGraph rather than my code. - [x] The bug is not resolved by updating to the latest stable version of LangGraph (or the specific integration package). -…### Checked other resources - [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 LangGraph rather than my code. - [x] The bug is not resolved by updating to the latest stable version of LangGraph (or the specific integration package). -…### Checked other resources - [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 LangGraph rather than my code. - [x] The bug is not resolved by updating to the latest stable version of LangGraph (or the specific integration package). -…### 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 my code. - [x] The bug is not resolved by updating to the latest stable version of LangChain (or the specific integration package). - [x]…### Component Other (Node CLI) ### Description #### Summary The Node CLI's API-key validation races `backend.ping()` against a five-second timeout. When the ping resolves first, the losing timer is not cleared, so a one-shot CLI process can remain alive until that timer expires. #### Steps to Reproduce Run the equivalent successful race from the current implementation: ```bash /usr/bin/time -f 'elapsed=%e…### 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 my code. - [x] The bug is not resolved by updating to the latest stable version of LangChain (or the specific integration package). - [x]…## Summary I'd like to add a small `BilibiliHotTool` (~130 LOC) that wraps Bilibili's public `/x/web-interface/popular` API, so smolagents Agents can fetch trending Chinese videos without writing one-off HTTP calls. ## Motivation smolagents' built-in web tools (`WebSearchTool`, `VisitWebpageTool`, `WikipediaSearchTool`) are all English-centric. As a Chinese content creator using smolagents for trend research, I…### 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 my code. - [x] The bug is not resolved by updating to the latest stable version of LangChain (or the specific integration package). - [x]…### 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 my code. - [x] The bug is not resolved by updating to the latest stable version of LangChain (or the specific integration package). - [x]…**Is your feature request related to a problem? Please describe.** Adding a BM25 sparse vector to a collection that already holds data requires reading every point out of Qdrant and writing it back, purely so Qdrant can tokenize a string it already stores. Since 1.18, we can add a named vector to an existing collection without recreating it, and the new vector is queryable immediately. It just returns nothing,…### 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 my code. - [x] The bug is not resolved by updating to the latest stable version of LangChain (or the specific integration package). - [x]…### Checked other resources - [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 LangGraph rather than my code. - [x] The bug is not resolved by updating to the latest stable version of LangGraph (or the specific integration package). -…It would be great if the backend can support [llama.cpp](https://github.com/ggml-org/llama.cpp) inference engines which would enable rapid expansion to new quantized model. For intance, unsloth released a new dynamic quantization method that works great and still gives you a usable model for a fraction of the resources the full one needs. Read [here](https://unsloth.ai/docs/basics/dynamic-3.0-ggufs) Which also…### Checked other resources - [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 LangGraph rather than my code. - [x] The bug is not resolved by updating to the latest stable version of LangGraph (or the specific integration package). -…The PR #3934 drafted a first declarative backend, `IWorkPagesDocumentBackend`, to parse Apple's iWork Pages documents. The backend needs some improvements to pair the capabilities of similar backend parsers like `MsWordDocumentBackend` for Word documents. Just drop a line if you're interested in resolving this issue. Make sure to follow the…Hello, I am trying to understand how to get a clean installation without unneccessary packages into my virtualenv. I have a setup with 2 servers: A is with AMD CPU to run docling B is an Nvidia Jetson ORIN running llama.cpp connected via LAN. During the installation of docling, pip installs a lot of dependencies including some packages that will never be able to run onto server A: pip install docling ``` ...…From: YRA Tech (document-accessibility tooling; Docling is the structural backbone of our conversion pipeline). Prior upstream work: docling#3803 (supersample), docling-parse#299/#300/#301/#314, the MLX conversion kit for docling#3943. Summary. We convert documents into tagged, PDF/UA-conformant PDFs using DoclingDocument as our canonical model. Docling's layout, table, and reading-order components are the…We are moving from Qdrant as a data store to Postgres with BM25 and Vector search plugins. The core implementation of the migration is here https://gitlab.com/TIBHannover/orkg/orkg-ask/core-data-migrationThe limits are so strict that sometimes I can't use them for 30 minutes or more.Tasks for implementing LLM selection: ## Model selection - Make the model selectable per request on the LLM endpoints (extraction and synthesis), with a default preserving current behavior. The selection query parameter is optional (e.g. `?model=tib:model-7b`, or `?model=openrouter:openai/gpt-5.4-mini`). - Support hosted models via OpenRouter (API-key based), in addition to the existing local/self-hosted…I possess some Ogham letters, scanned as monochromatic TIFFs. I want to convert them to UTF.I would rather see the original pixels than a perfectly rendered but mathematically incorrect digital word.I massively prefer inline code completion to agentic workflows, mostly due to security, reliability and auditability.The upstream Opencode repository and therefore its path to the container image changed. As the name change isn't automatically respected during the Dockerfile building, it fails. To reproduce: 1. Open https://github.com/sst/opencode in your browser 2. See you're redirected to https://github.com/anomalyco/opencode The Dockerfile needs to be changed to the new name in order for the builds to succeed.Provide transcript or subtitles of the voice in the video, both within the highlights and outside the highlights.I am setting up an ollama server. I installed xrdp using: sudo apt-get update sudo apt-get install xrdp This works fine when I do not have the Nvidia card installed. But when I have the card in and remote using RDP from Windows, I get a connection and then the window closes quickly. The Nvidia P40 works fine b/c it is detected by nvidia-smi, and am able to see its TPU, temp and memory utilization increase when I…Hi, I gave it a try. and started wally. I do not know if this is a bug or not, but it looks like a bug. One other thing I noticed was. It was pointed out it is a namespace filter. Meaning I can run it on a single project as a starter. That is nothing to worry though. ``` ENV WALLY_GITLAB_TOKEN= ENV WALLY_GITLAB_URL=https://gitlab.com/ ENV WALLY_DRY_RUN=true ENV WALLY_LOG_LEVEL=debug ENV…My Fedora system recently crashes with strange error messages, so I am worried that my ~5 years old SSD is failing. Now I want to rescue my data from this SSD and reinstall Fedora from scratch on a new SSD. I created a bootable Fedora USB flash drive and booted it without mounting the failing SSD. Now I want to rescue the data from this SSD onto an external 8TB HDD (~4TB free space). The SSD has a capacity of…Is there a FOSS alternative to GitHub Copilot, preferably one that can be integrated into Vim?Excel document I have the following XLSX-formatted spreadsheet in Google Sheets I'm trying to set a formula on cell R5 (colored in red) so that it adds up all the times in the range L4:N42 but whatever I try it I either get 00:00:00 or all kinds of errors, not even chatGPT can help. Current Formula =ARRAY_CONSTRAIN(ARRAYFORMULA(SUM(IF(ISNUMBER(L4:N42);L4:N42;0)));1;1)Excel function where if any row values in column A include a keyword from the list in column D but does not include a keyword from the list in column E, write True, otherwise, write nothing/keep it blank. Attempts: =IF(SUMPRODUCT(--(NOT(ISERR(SEARCH($D$2:$D$3, A2)))) * (ISERR(SEARCH($E$2:$E$3, A2)))) > 0, "True", "") This does not work but if the Exclude is singular and not a list, it does:…Mock all the engines' components and write the LLM endpoints test casesHow can I get the transcript of a YouTube Short? I don't see any option to download it.<h2>Scenario:</h2>
<p>I have a master Google Sheets spreadsheet with dozens of pages linking to Data Studio.</p>
<p>Each page has an IMPORTRANGE formula linking to other Google Sheets spreadsheets.</p>
<p>These other spreadsheets have multiple pages and are created monthly, meaning a new URL for each spreadsheet.</p>
<h2>I want to:</h2>
<p>Have one master spreadsheet to avoid re-linking to Data Studio for each…UPDATE Now it only need to scan the whole row instead of only the cell next to it. I am quite new to google sheets and I want to achieve the following: Lets say we have list 1: Red apple yellow banana orange orange juicy melon We also have list 2 that looks as follows: giant red apple yellow banana orange orange purple water melon Now I want to write a function where I loop through each item in List 2 where I…I have a thousand handwritten responses to a paper form and need to export all the data in a table.I'm wondering if someone can help me. I have a Google sheets workbook with more than 50 individual sheets. This workbook has a primary 'index' page that collates set data fields from each sheet. I'm currently expanding this but the formula I am using requires me to change the referencing sheet per row so the data is pulled for the correct sheet. My questions is, is there a way to automate this process as it is…I am trying to query the highest and lowest scoring colour within a group. Then to query what the benchmark score is for those colours in all groups. I'm totally stuck on an approach. I have marked in red in the 2nd tab the expected values: https://docs.google.com/spreadsheets/d/1H7Ighd3fHUc3fJgAXWiqx_Ptl51OvZwXMMToczYnx8A/edit?usp=sharingHow can Google Sheets take a range of multiple columns and rows and stack or consolidate all the non-empty cells into one column? For this case, I want non-unique cells. Also, each cell in the range is a string that has commas and spaces so if needed, a special char will need to be as a delimiter. There are several questions already that are about concatenating columns, but they are all concatenating the values…Is there any good free Hebrew OCR for Linux? At least something trainable would help.I am looking for an interactive transcript API that is available for free/open source and usable for commercial websites.