2026-06-21 - 2026-09-21
Overview
14 Pull requests merged by 1 user
Merged
#54 Fix/omniroute pr agent timeout
Merged
#53 fix(update.sh): stop config sync loop from dying silently on a missing key
Merged
#52 Add Qdrant/Neo4j RAG storage + update.sh gum fixes
Merged
#50 feat: add qdrant and neo4j for RAG vector/graph storage
Merged
#51 feat: remove llama-server-fast (Qwen3-4B classifier model)
Merged
#49 Fix llama-server-fast context-size exhaustion breaking Auto Mode
Merged
#48 Run git pull first in update.sh, not mid-script
Merged
#47 Fix GPU pinned at 100% with two containers, flaky render group
Merged
#46 Downloader for Qwen-Image weights, switch-model.sh
Merged
#45 Add llama-server-fast: small non-thinking classifier/fast model
Merged
#30 feat(llama.cpp): raise default context to 128K, document RAM/SSD offload knobs
Merged
#20 fix(litellm): default max_tokens=4096 for the reasoning model
Merged
#19 feat(litellm): add UI_USERNAME/UI_PASSWORD for the admin UI login
Merged
#18 fix(downloader): run as root to fix permission denied on models volume
31 Issues closed from 1 user
Closed
#5 Verify the stack on the real Radeon R9700 box (GPU passthrough + Lazytainer idle-stop)
Closed
#43 scripts/switch-model.sh: swap GPU residency between Qwen and ComfyUI
Closed
#42 Extend downloader for the diffusion model weights
Closed
#44 Add llama-server-fast: small non-thinking classifier/fast model for qwen-code Auto Mode
Closed
#39 Best diffusion model given full ~32GB headroom (Qwen stopped)
Closed
#40 Why lazytainer idle-stop doesn't work today, and how the switch script should handle it
Closed
#37 OmniRoute: docker-compose service design and secrets/config plan
Closed
#32 OmniRoute: can it route to arbitrary OpenAI-compatible local endpoints (llama-server), not just Ollama?
Closed
#36 OmniRoute: per-workload virtual keys, admin UI, and deployment shape (Docker/compose)
Closed
#35 OmniRoute: web-search tool equivalent to the SearXNG standalone endpoint
Closed
#33 OmniRoute due-diligence: maintainer, repo history, npm package trust
Closed
#34 OmniRoute: memory/knowledgebase parity with litellm-pgvector
Closed
#29 Point opencode's provider block at litellm.home with a virtual key
Closed
#28 Should opencode's context/output limits be generated from litellm-config.yaml, or hand-maintained?
Closed
#27 How does opencode's auto-compact actually work — trigger threshold and config surface?
Closed
#25 LangChain+pgvector direct RAG vs. the litellm-pgvector connector — which fits this stack?
Closed
#23 How does LiteLLM's knowledgebase/vector-store feature work, and what does it need?
Closed
#22 How does LiteLLM's SearXNG web-search integration work, and what does wiring it in require?
Closed
#15 Migrate Open WebUI (and existing consumers) to route through the new proxy
Closed
#14 Author the docker-compose service for the chosen proxy
Closed
#16 How should the proxy queue and prioritize concurrent requests across workloads?
Closed
#13 Network/hostname plan for exposing the proxy (LAN + external)
Closed
#12 How should per-workload API keys/accounts be provisioned, rotated, and documented?
Closed
#11 Reference cloud model/pricing for the shadow-cost estimate
Closed
#10 Which self-hosted AI gateway/proxy tool fits this effort's needs?
Closed
#8 How is ai.home / ai.haylan.ch reverse-proxied and TLS-terminated?
Closed
#6 Write local-usage docs for Claude Code CLI, Kimi CLI, and OpenCode CLI against the local endpoint
Closed
#7 Research OpenCode CLI installation and local-endpoint setup
Closed
#4 Author the docker-compose stack (llama.cpp + Open WebUI + Qdrant + Lazytainer)
Closed
#2 Confirm quantization availability for Qwen3.8-27B
Closed
#3 Confirm Qwen3.8-27B tool-calling compatibility with llama.cpp's Anthropic Messages API shim
40 Issues created by 1 user
Opened
#1 Local AI inference stack: llama.cpp + Qwen3.8-27B on Radeon R9700
Opened
#2 Confirm quantization availability for Qwen3.8-27B
Opened
#3 Confirm Qwen3.8-27B tool-calling compatibility with llama.cpp's Anthropic Messages API shim
Opened
#4 Author the docker-compose stack (llama.cpp + Open WebUI + Qdrant + Lazytainer)
Opened
#5 Verify the stack on the real Radeon R9700 box (GPU passthrough + Lazytainer idle-stop)
Opened
#6 Write local-usage docs for Claude Code CLI, Kimi CLI, and OpenCode CLI against the local endpoint
Opened
#7 Research OpenCode CLI installation and local-endpoint setup
Opened
#8 How is ai.home / ai.haylan.ch reverse-proxied and TLS-terminated?
Opened
#9 AI gateway/proxy: routing, per-workload keys, and cloud-cost estimation for the local LLM stack
Opened
#10 Which self-hosted AI gateway/proxy tool fits this effort's needs?
Opened
#11 Reference cloud model/pricing for the shadow-cost estimate
Opened
#12 How should per-workload API keys/accounts be provisioned, rotated, and documented?
Opened
#13 Network/hostname plan for exposing the proxy (LAN + external)
Opened
#14 Author the docker-compose service for the chosen proxy
Opened
#15 Migrate Open WebUI (and existing consumers) to route through the new proxy
Opened
#16 How should the proxy queue and prioritize concurrent requests across workloads?
Opened
#17 Verify the AI proxy stack on the real R9700 box (LiteLLM scheduler smoke test)
Opened
#21 LiteLLM gateway: search, knowledgebase, and gateway-level memory
Opened
#22 How does LiteLLM's SearXNG web-search integration work, and what does wiring it in require?
Opened
#23 How does LiteLLM's knowledgebase/vector-store feature work, and what does it need?
Opened
#24 Verify search + knowledgebase wiring (#21) on the real R9700 box
Opened
#25 LangChain+pgvector direct RAG vs. the litellm-pgvector connector — which fits this stack?
Opened
#26 opencode: context size and compaction settings for the litellm/qwen3.8-27b-local stack
Opened
#27 How does opencode's auto-compact actually work — trigger threshold and config surface?
Opened
#28 Should opencode's context/output limits be generated from litellm-config.yaml, or hand-maintained?
Opened
#29 Point opencode's provider block at litellm.home with a virtual key
Opened
#31 Migrate the AI gateway from LiteLLM to OmniRoute
Opened
#32 OmniRoute: can it route to arbitrary OpenAI-compatible local endpoints (llama-server), not just Ollama?
Opened
#33 OmniRoute due-diligence: maintainer, repo history, npm package trust
Opened
#34 OmniRoute: memory/knowledgebase parity with litellm-pgvector
Opened
#35 OmniRoute: web-search tool equivalent to the SearXNG standalone endpoint
Opened
#36 OmniRoute: per-workload virtual keys, admin UI, and deployment shape (Docker/compose)
Opened
#37 OmniRoute: docker-compose service design and secrets/config plan
Opened
#38 Local image generation via ComfyUI + OmniRoute
Opened
#39 Best diffusion model given full ~32GB headroom (Qwen stopped)
Opened
#40 Why lazytainer idle-stop doesn't work today, and how the switch script should handle it
Opened
#41 Add comfyui service to docker-compose.yml, register with OmniRoute
Opened
#42 Extend downloader for the diffusion model weights
Opened
#43 scripts/switch-model.sh: swap GPU residency between Qwen and ComfyUI
Opened
#44 Add llama-server-fast: small non-thinking classifier/fast model for qwen-code Auto Mode