Repository navigation
What's Changed
Bug Fixes & Code Quality
- Fixed unsafe memory zeroing in otl-rag tests
- Resolved all dead_code warnings across workspace
- Fixed clippy warnings (uninlined_format_args, for_kv_map, excessive_nesting)
- Refactored nested async code in cache module
Kubernetes Deployment
- Fixed DATABASE_URL configuration in k8s configmap
- Fixed PostgreSQL credentials in connection string
- Updated deployment image tag to v0.2.1
vLLM/GPU Integration
- Configured vLLM via OpenAI-compatible API
- Support for external GPU server inference
- Ollama embedding provider on GPU server
- Updated vector dimension for nomic-embed-text (768)
Configuration
vLLM Setup
```yaml
LLM_PROVIDER: "openai"
LLM_MODEL: "Qwen/Qwen2.5-7B-Instruct"
OPENAI_API_BASE: "http://:8000/v1"
EMBEDDING_PROVIDER: "ollama"
EMBEDDING_MODEL: "nomic-embed-text"
```
Commits
- docs: add v0.2.1 release session log (4bed5c3)
- chore: update k8s config to use vLLM via OpenAI-compatible API (4a6b661)
- fix: add DATABASE_URL and correct PostgreSQL credentials in k8s configmap (fd41540)
- chore: update image tag to v0.2.1 (6407c72)
- fix: resolve dead_code warnings in otl-cli (ca2101f)
- fix: apply clippy suggestions across workspace (933d15f)
- fix: remove unused code in otl-extractor (b82ca60)
- fix: remove unused calculate_confidence method in otl-rag (5948d1c)
- fix: remove unsafe memory zeroing in keyword search tests (9a04f6a)
- feat: add separate embedding provider configuration (e9ed22a)
- feat: add vLLM/OpenAI-compatible API support via OPENAI_API_BASE (a488564)
Docker Image
```bash
docker pull ghcr.io/hephaex/otl-api:v0.2.1
```