Issues

• flm serve: a client that resets before accept stops the server accepting connections (HttpSession ctor throws ENOTCONN)

  ROCm/FastFlowLM#762 · Filter by author noamsto opened on Sep 28, 2026
  1

• flm serve error responses: sent with HTTP 200, include raw exception text, and GET /api/ps is undefined behaviour with no model loaded

  ROCm/FastFlowLM#761 · Filter by author noamsto opened on Sep 28, 2026

• flm serve: /v1/completions, /api/generate and /api/chat keep decoding after the client disconnects

  ROCm/FastFlowLM#760 · Filter by author noamsto opened on Sep 28, 2026
  1

• flm serve: /v1/completions and other routes run outside the NPU queue and race with model swaps

  ROCm/FastFlowLM#759 · Filter by author noamsto opened on Sep 28, 2026
  1

• /v1/embeddings ignores prompt_name / task_type, so every input is embedded as a query

  ROCm/FastFlowLM#758 · Filter by author noamsto opened on Sep 28, 2026

• flm serve: streaming /api/chat segfaults the server on every request (insert() twice, generate() never called)

  ROCm/FastFlowLM#757 · Filter by author noamsto opened on Sep 28, 2026

• flm serve: a request body with invalid UTF-8 leaves the NPU locked, and every later request queues forever

  ROCm/FastFlowLM#756 · Filter by author noamsto opened on Sep 28, 2026

• flm serve crashes when a request is missing a required field (POST {})

  ROCm/FastFlowLM#755 · Filter by author noamsto opened on Sep 28, 2026

• [Bug][Linux] /v1/embeddings always fails with qds_device::wait() unexpected command state (Krackan Point, FLM 1.0.6)

  ROCm/FastFlowLM#752 · Filter by author ripiox opened on Sep 23, 2026

• Tool calling serializes boolean arguments as strings with Qwen3.5 ("True" instead of true)

  ROCm/FastFlowLM#750 · Filter by author wjaschenkov opened on Sep 23, 2026

• flm serve OpenAI-compatible endpoint hardcodes max_tokens default = 4096 → long responses are silently truncated (finish_reason: "length") when the client omits max_tokens

  ROCm/FastFlowLM#749 · Filter by author camilo6castell opened on Sep 22, 2026

• [Bug, Issue]: ASR model (whisper-large-v3-turbo) not explicitly named in v1/models, so doesn't work with stricter openAI compatible connections

  ROCm/FastFlowLM#746 · Filter by author Seyij opened on Sep 19, 2026
  1

• [Feature]: Add support for Cactus Compute Needle model

  ROCm/FastFlowLM#745 · Filter by author julianxhokaxhiu opened on Sep 18, 2026

• No valid checkpoint to restore + Max length reached on multi-turn tool calls (Qwen3.6-MoE, NPU)

  ROCm/FastFlowLM#744 · Filter by author winaes13-ux opened on Sep 18, 2026
  2

• Truncated tool call is emitted as a complete one (unterminated heredoc reaches the client)

  ROCm/FastFlowLM#741 · Filter by author Javinator9889 opened on Sep 18, 2026
  1

• Non-streaming responses leak <think> tags into content for qwen3.5, qwen3.6-moe, qwen3vl and qwen3.5-omni

  ROCm/FastFlowLM#739 · Filter by author Javinator9889 opened on Sep 18, 2026
  1

• [Feature]: KV-cache reuse for shared-prefix one-shot requests (prefix caching / partial truncate)

  ROCm/FastFlowLM#737 · Filter by author thunderbug1 opened on Sep 17, 2026

• Prompt cache: identical resend never hits, partial match discards everything, and nothing is sent during prefill

  ROCm/FastFlowLM#733 · Filter by author Javinator9889 opened on Sep 17, 2026
  3

• v1.0.4 regression: embed-gemma:300m embeddings via --embed 1 are wrong (uncorrelated with v0.9.39 on identical weights)

  ROCm/FastFlowLM#732 · Filter by author GhostInTheNPU opened on Sep 17, 2026

• [Feature/Model Request]: Support for NPU-accelerated Text-to-Speech (TTS) models

  ROCm/FastFlowLM#731 · Filter by author Seyij opened on Sep 17, 2026

• [Documentation]: How to generate SRT/VTT transcripts

  ROCm/FastFlowLM#726 · Filter by author vincent-sparks opened on Sep 13, 2026

• [Issue]: [Bug][Linux] flm serve crash: C++ assertion failure '!this->empty()' in std::vector<logits_t>::back()

  ROCm/FastFlowLM#725 · Filter by author austrasien opened on Sep 12, 2026
  1

• [Issue]: Gemma4 tool-call parsing produces wrong name and arguments (v1.0.5)

  ROCm/FastFlowLM#722 · Filter by author skarpov2 opened on Sep 10, 2026
  3

• Flm with gemma 4 has very poor ability to call tools

  ROCm/FastFlowLM#721 · Filter by author tidotua opened on Sep 10, 2026
  4

• Inconsistent memory consumption?

  ROCm/FastFlowLM#720 · Filter by author VladimirVLF opened on Sep 10, 2026