Running Qwen 27B locally via MCP + Docker sandbox to refactor shell scripts
ag789 · reddit · 2026-10-27
A Reddit user details using a locally-run Qwen 27B model, connected through a custom MCP server and Docker sandbox, to refactor a complex shell script for llama.cpp. The model caught subtle details unprompted (filtering config vars not meant for llama-server) and accurately verified manual edits hunk-by-hunk from a unified diff. Main complaints: very slow inference and rapidly exhausting the 65,535 context budget in long sessions. Full verification dialogue and linked GitHub projects included.
More from coding & agent
- figwig format ships with CLI tools and an agentic skill for bioinformatics sparse data — jmschreiber91 · 2026-10-28
- Harmonic launches Aristotle, an AI theorem prover that formally verifies C and Verilog code — xennygrimmato_ · 2026-10-28
- Designers argue personal agents must build user-fitting into the product, not leave it as a switch — sujingshen · 2026-10-28
- Nous Research Revamps Hermes Agent Plans: Cloud Agent Included, Daily/Weekly Limits Replace Monthly Credits — NousResearch · 2026-10-28
- Claude Code can now edit videos and cut TikTok-ready clips — 8 prompts to try — aitrendz_xyz · 2026-10-28
- Harmonic Launches Aristotle for Automated Formal Verification of C and Verilog Code — satnam6502 · 2026-10-28