Skip to content

feat: add Candle local inference + cross-platform CI + updated benchm… - #6

Merged
Nitin-100 merged 1 commit into
masterfrom
feature/local-inference
Feb 24, 2026
Merged

Nitin-100 merged 1 commit into
masterfrom
feature/local-inference

Conversation

@Nitin-100

Copy link
Copy Markdown
Owner

…arks

  • Add local inference via Candle (Rust-native ML) behind --features local flag
  • 8 model presets: tinyllama, smollm (135m/360m/1.7b), phi2, qwen (0.5b/1.5b)
  • CLI flags: --local and --model for both agent and gateway commands
  • Air-gapped deployment support (models cached after first download)
  • 4 chat templates: TinyLlama, ChatML, Llama2, Llama3

CI/CD:

  • Add release.yml: cross-platform release builds (Linux x86_64/ARM64, macOS x86_64/ARM64, Windows x86_64) with GitHub Release automation
  • Add local.yml: dedicated CI for local inference feature (build, test, clippy, gateway smoke tests on all 3 platforms)
  • Update ci.yml: matrix now tests both standard and --features local builds, clippy checks both feature sets
  • Update bench.yml: trigger on feature/local-inference branch

README:

  • Update benchmark numbers: 6.68 MB idle (was 7.02), 3,774 req/s bulk
  • Add local inference benchmark table: 9.34 MB idle, 3,906 req/s bulk
  • Add comprehensive local model docs: build, configure, test, air-gapped
  • Add CI badges for local.yml and release.yml workflows
  • Update test count to 474

Binary sizes: 4.21 MB standard, 7.79 MB with local (+3.58 MB) 474 tests pass, 0 clippy warnings, 0 fmt diffs

…arks

- Add local inference via Candle (Rust-native ML) behind --features local flag
- 8 model presets: tinyllama, smollm (135m/360m/1.7b), phi2, qwen (0.5b/1.5b)
- CLI flags: --local and --model for both agent and gateway commands
- Air-gapped deployment support (models cached after first download)
- 4 chat templates: TinyLlama, ChatML, Llama2, Llama3

CI/CD:
- Add release.yml: cross-platform release builds (Linux x86_64/ARM64,
  macOS x86_64/ARM64, Windows x86_64) with GitHub Release automation
- Add local.yml: dedicated CI for local inference feature (build, test,
  clippy, gateway smoke tests on all 3 platforms)
- Update ci.yml: matrix now tests both standard and --features local builds,
  clippy checks both feature sets
- Update bench.yml: trigger on feature/local-inference branch

README:
- Update benchmark numbers: 6.68 MB idle (was 7.02), 3,774 req/s bulk
- Add local inference benchmark table: 9.34 MB idle, 3,906 req/s bulk
- Add comprehensive local model docs: build, configure, test, air-gapped
- Add CI badges for local.yml and release.yml workflows
- Update test count to 474

Binary sizes: 4.21 MB standard, 7.79 MB with local (+3.58 MB)
474 tests pass, 0 clippy warnings, 0 fmt diffs
@Nitin-100
Nitin-100 merged commit 2381620 into master Feb 24, 2026
30 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant