EVEngine

1,995 Commits 1 Branch 0 Tags

3 Commits

Include renames

Author	SHA1	Message	Date
西风逍遥游	811eff1c3a	ci(quality): warnings-as-errors gate, sanitizers, smoke, Python lint/tests - EVENGINE_ENABLE_STRICT_WARNINGS option: MSVC /W4 /WX, GCC/Clang -Wall -Wextra -Werror, with documented noise suppressions and -isystem/-external:I for third-party headers; enabled on the Windows/Linux debug CI jobs. - Fix warnings across engine, modules and tests exposed by the gate, including real bugs (incomplete-type deletes, Timespan microsecond overflow, overload hiding, enum switch gaps, dead code). - Add Linux ASan+UBSan job, example smoke step, python-tools test job, ruff config (scripts+tools), weekly Linux coverage job and a CodeQL workflow.	2026-08-22 02:27:37 +08:00
西风逍遥游	ea92c34dab	tools(vision-prefilter): switch to llama.cpp llama-server backend Drop the heavy transformers/torch/bitsandbytes inference path in favour of llama.cpp llama-server serving a Qwen2-VL-2B GGUF. The service becomes a thin FastAPI gateway that forwards to the OpenAI-compatible /v1/chat/completions API. - Add GBNF grammar (grammar.py) that enforces the standard JSON at token sampling level, making free-text/markdown output impossible - model.py is now an HTTP client (LlamaServer) over llama-server; no torch - server.py/__main__.py talk to VISION_PREFILTER_BACKEND (default :8080) - requirements.txt now only fastapi/uvicorn/pydantic/requests - Launchers start llama-server + gateway; add download_model.sh (Q4_K_M GGUF) - README rewritten for the llama.cpp deployment; tests still pass	2026-08-16 15:18:07 +08:00
西风逍遥游	d8a77298ec	tools(vision-prefilter): local Qwen2-VL-2B 4bit vision pre-filter service Low-cost local VLM HTTP service that batches render snapshots + compact geometry text and returns a strict JSON risk rating, filtering safe viewpoints before an expensive high-precision VLM is called. - Qwen2-VL-2B 4bit (bitsandbytes nf4) lazy-loaded wrapper with JSON-only output enforcement and robust parsing (no free-text leak on failure) - FastAPI /v1/prefilter/batch + /health + /protocol endpoints - Standardised pydantic input/output protocol (risk_score 0-3, problem regions with pixel bbox, need_high_precision_review) - Deterministic business rules: road occlusion=high, vegetation cluster= medium, balanced layout=no risk - Background launchers (sh/bat) + batch client example + no-torch tests	2026-08-16 15:10:36 +08:00

Author

SHA1

Message

Date

西风逍遥游

811eff1c3a

ci(quality): warnings-as-errors gate, sanitizers, smoke, Python lint/tests

- EVENGINE_ENABLE_STRICT_WARNINGS option: MSVC /W4 /WX, GCC/Clang
  -Wall -Wextra -Werror, with documented noise suppressions and
  -isystem/-external:I for third-party headers; enabled on the
  Windows/Linux debug CI jobs.
- Fix warnings across engine, modules and tests exposed by the gate,
  including real bugs (incomplete-type deletes, Timespan microsecond
  overflow, overload hiding, enum switch gaps, dead code).
- Add Linux ASan+UBSan job, example smoke step, python-tools test
  job, ruff config (scripts+tools), weekly Linux coverage job and a
  CodeQL workflow.

2026-08-22 02:27:37 +08:00

西风逍遥游

ea92c34dab

tools(vision-prefilter): switch to llama.cpp llama-server backend

Drop the heavy transformers/torch/bitsandbytes inference path in favour of
llama.cpp llama-server serving a Qwen2-VL-2B GGUF. The service becomes a thin
FastAPI gateway that forwards to the OpenAI-compatible /v1/chat/completions API.

- Add GBNF grammar (grammar.py) that enforces the standard JSON at token
  sampling level, making free-text/markdown output impossible
- model.py is now an HTTP client (LlamaServer) over llama-server; no torch
- server.py/__main__.py talk to VISION_PREFILTER_BACKEND (default :8080)
- requirements.txt now only fastapi/uvicorn/pydantic/requests
- Launchers start llama-server + gateway; add download_model.sh (Q4_K_M GGUF)
- README rewritten for the llama.cpp deployment; tests still pass

2026-08-16 15:18:07 +08:00

西风逍遥游

d8a77298ec

tools(vision-prefilter): local Qwen2-VL-2B 4bit vision pre-filter service

Low-cost local VLM HTTP service that batches render snapshots + compact
geometry text and returns a strict JSON risk rating, filtering safe
viewpoints before an expensive high-precision VLM is called.

- Qwen2-VL-2B 4bit (bitsandbytes nf4) lazy-loaded wrapper with JSON-only
  output enforcement and robust parsing (no free-text leak on failure)
- FastAPI /v1/prefilter/batch + /health + /protocol endpoints
- Standardised pydantic input/output protocol (risk_score 0-3, problem
  regions with pixel bbox, need_high_precision_review)
- Deterministic business rules: road occlusion=high, vegetation cluster=
  medium, balanced layout=no risk
- Background launchers (sh/bat) + batch client example + no-torch tests

2026-08-16 15:10:36 +08:00