Repository navigation
Releases
Releases · roryeckel/sector-ai
v1.5.0
Highlights
- OpenAI-compatible backend.
SECTOR_AI_LLM_PROVIDER=openaisends Chat Completions to any/v1server (Ollama, LM Studio, vLLM, llama.cpp, OpenRouter, OpenAI, self-hosted). Context length is detected from the server,/tokensreports the real context, and summary compaction replaced the fixed message cutoff. Ollama remains the default. (#12) - Markdown replies. Chat and vision answers render as Telegram formatting — bold, italic, inline and fenced code, links, block quotes, spoilers, list markers. A rejected formatted edit falls back to plain text, and replies over 4096 characters are split across messages instead of failing.
SECTOR_AI_MARKDOWN=falserestores raw text. (#14) - Release pipeline. A three-stage release flow (prepare → draft → publish) with a repo-level skill so agents follow the same path. (#13)
Full Changelog: v1.4.1...v1.5.0
v1.4.1
What's Changed
Other changes
- Update all pip packages to latest versions and add development tools (ruff/pytest) by @roryeckel in #2
- Upgrade dependencies to latest, fix stream edits, and add a full test suite by @roryeckel in #3
- Replace the JSON config file with typed, environment-driven settings by @roryeckel in #4
- Accept the documented null context length from the environment by @roryeckel in #6
- Publish to PyPI and GHCR through a staged release pipeline by @roryeckel in #5
- Run model calls without blocking the event loop by @roryeckel in #8
- Redact the bot token from log output by @roryeckel in #10
New Contributors
- @roryeckel made their first contribution in #3
- @github-actions[bot] made their first contribution in #7
Full Changelog: https://github.com/roryeckel/sector-ai/commits/v1.4.1