Skip to content

Releases: zenithfoundry/slm-gate

v1.3.0

Choose a tag to compare

@zenithfoundry-slm-gate-release zenithfoundry-slm-gate-release released this 27 Sep 02:23
3a23d61

1.3.0 (2026-09-27)

Features

  • setup: AI setup prompt, slm-gate --version, and switching from a git checkout (429c362)
  • setup: connect-first tool pages, init points to the AI setup prompt, npm badges; fix setup errors in tool docs (1095495)

Bug Fixes

  • doctor: skip the file check for a toolbox started by a command such as npx (e9b548e)

v1.2.2

Choose a tag to compare

@zenithfoundry-slm-gate-release zenithfoundry-slm-gate-release released this 26 Sep 22:10
d7eac76

1.2.2 (2026-09-26)

Bug Fixes

  • deps: bump actions/setup-node from 4.4.0 to 7.0.0 (2d473b9)
  • deps: bump actions/setup-node from 4.4.0 to 7.0.0 (d54083a)
  • deps: bump actions/upload-pages-artifact from 3.0.1 to 5.0.0 (e891e1b)
  • deps: bump actions/upload-pages-artifact from 3.0.1 to 5.0.0 (cec57e1)
  • deps: bump dotenv from 16.6.1 to 18.0.3 (9217084)
  • deps: bump pnpm/action-setup from 3.0.0 to 6.1.0 (ecd63c5)
  • deps: bump pnpm/action-setup from 3.0.0 to 6.1.0 (1e8d7d3)

v1.2.1

Choose a tag to compare

@zenithfoundry-slm-gate-release zenithfoundry-slm-gate-release released this 26 Sep 21:49
42e7051

1.2.1 (2026-09-26)

Bug Fixes

  • deps: bump the minor-and-patch group across 1 directory with 9 updates (8e3ff27)
  • deps: bump the minor-and-patch group across 1 directory with 9 updates (df0e0fa)

v1.2.0

Choose a tag to compare

@zenithfoundry-slm-gate-release zenithfoundry-slm-gate-release released this 26 Sep 20:28
fbfe8d4

1.2.0 (2026-09-26)

Features

  • cli: add init and mcp commands for npm installs; accept the 64 and 128 GB RAM presets (f85f958)
  • config: RAM presets from 16 to 128 GB (new 48 GB), custom above 128 GB; drop under-16 GB; align docs, point to llmfit (aceed2a)

v1.1.0

Choose a tag to compare

@zenithfoundry-slm-gate-release zenithfoundry-slm-gate-release released this 26 Sep 19:34
a029169

1.1.0 (2026-09-26)

Features

  • config: keep settings and data in ~/.slm-gate for npm installs; give real-gate tests their own ports (236f8f5)
  • config: keep settings and data in ~/.slm-gate for npm installs; give real-gate tests their own ports (32c0465)

v1.0.0

Choose a tag to compare

@github-actions github-actions released this 26 Sep 16:23
f25c434

1.0.0 (2026-09-26)

Features

  • add agent column to events table and update cycle rate calculation to use provider-specific token metrics (a117e5a)
  • add cycle window extension metrics to report, sync, and dashboard configuration (6f5d3b6)
  • add cycle_minutes_saved metrics for ChatGPT, Claude, and Gemini to sync configuration (ae5756e)
  • add desktop notifications for slm-gate issues and implement startup checks (c936a93)
  • add Google Gemini support and improve dynamic API URL path construction (70d3b3d)
  • add Langfuse synchronization and local ledger reset utilities to CLI (74d4c4b)
  • add OLLAMA_KEEP_ALIVE configuration to control model memory duration (c9b2afe)
  • add prompt caching support to Anthropic system, tool, and message formats (a143e50)
  • add PROVIDER configuration and improve ledger sync logging, error handling, and provider attribution logic. (0f34436)
  • add provider tracking to ledger events with schema migration and utility functions for analytics (de740ac)
  • add RAM-preset configuration templates and expand documentation for LLM-gate deployment. add cli and docs (8b7ad06)
  • add reporting and verification for local ledger against Langfuse (e8c4ccc)
  • add SLM_GATE_TESTING_MODEL config to support specialized benchmarking models (ced6b05)
  • add spike passthrough proxy and llm-gate forwarding logic (61ced9f)
  • add test:summary script to package.json for concise test output fix tests (0a934d6)
  • add TLS downstream E2E test and refactor Langfuse integration to use direct SDK client (6bb566f)
  • add transient problem handling to startup checks and model gate notifications (6b5c46b)
  • dashboard: add static site export and local server for metrics dashboard (c6cab0b)
  • dashboard: enhance widget placement logic and add verification for persisted placements (cc9e4dd)
  • dashboard: update README and add dashboard publish script (df6a176)
  • docs: enhance README with detailed visual explanations and diagrams (4428427)
  • docs: update README to clarify MCP and model gate functionalities (5bad707)
  • doctor: apple-silicon-aware hardware checks + fallback config (b58489f)
  • enforce local-only access for slm-gate and mcp-gate servers (998dee7)
  • enforce Native Structured Outputs across all model interactions to improve stability and reliability (005e4ab)
  • enhance CI workflow with better-sqlite3 verification and update package engines (61ab38b)
  • enhance Ollama probing with detailed error handling and testing scenarios (353bf5e)
  • enhance telemetry transport with environment tagging and implement shared exponential backoff for retries (6efdae8)
  • enhance toolbox result handling to preserve all content while conditioning only text (07d9f5a)
  • implement adaptive distillation feedback loop with semantic embedding support and user-configurable preservation patterns (b5cd027)
  • implement ambiguity resolution pipeline with decision extraction, evidence-based reasoning, and risk-aware gating (7ccde4c)
  • implement dynamic ledger path creation for E2E tests and improve environment variable handling (61ab38b)
  • implement exponential backoff with jitter for network/rate-limit errors and update classification prompt and test harness logging (c72ad01)
  • implement MCP gateway server with proxy capabilities and prompt conditioning pipeline (1c8f74b)
  • implement monotonic SLM response verification logic and add unit tests (8f3678f)
  • implement native LLM gateway server with request routing, multi-format proxying, and SLM integration. (52cea7e)
  • implement persistent Langfuse queue with sync utility and telemetry metrics support (7bc4c49)
  • implement semantic caching for LLM responses using vector embeddings and file context verification (25e62a1)
  • implement SLM integration with Ollama/OpenAI support, self-consistency reasoning, and classification utilities. (64c7720)
  • implement sqlite-based event ledger with langfuse mirroring and cost calculation logic (ffe2186)
  • implement stranded server detection and cleanup for MCP servers (7505c0f)
  • implement subscription-based plan management for token rate limiting and update configuration examples. (bab2d4b)
  • implement text distillation module with configurable pattern preservation for model optimization (4cb170d)
  • implement TLS analytics mapping and downstream configuration in tech-lead-stack adapter (d2293dc)
  • initialize project structure with core gate middleware, configuration management, and architecture documentation (093d1c1)
  • introduce E2E test suite for ambiguity resolution and standalone MCP, and add logging to SLM client initialization and request pipelines. (de0b762)
  • introduce preservation patterns for SLM contextual compression (fdb584a)
  • llm-gate: implement local-first request handling and response logging (ea05b74)
  • pricing: update model pricing and introduce provider registry (7251b92)
  • support Anthropic-style API authentication and URL formatting in doctor check (a474cda)
  • tests: reorganize imports and enhance cycle metrics tests with additional utility functions (98d8927)
  • update preserve token format to brackets and add robust fallback restoration logic (62611fb)

B...

Read more