This release keeps dependencies and maintenance posture current for teams operating this tool.
✓ No known CVEs patched in this version
Topics
+2 more
Summary
AI summaryMinor fixes and improvements.
Changes in this release
| Type | Severity | Summary | CVE |
|---|---|---|---|
| Refactor | Low |
Refactored chat loop into unified _run_native_chat driver with shared machinery. Refactored chat loop into unified _run_native_chat driver with shared machinery. Source: llm_adapter@2026-06-07 Confidence: high |
— |
| Refactor | Low |
Extracted session persistence into dedicated SessionStore; Kodiqa retains thin wrappers. Extracted session persistence into dedicated SessionStore; Kodiqa retains thin wrappers. Source: llm_adapter@2026-06-07 Confidence: high |
— |
Full changelog
Internal refactor — no user-facing behavior change.
Changed
- Unified chat loop. The two near-identical native tool-calling loops (Claude and all OpenAI-compatible providers, ~180 lines each, ~90% shared) are now one `_run_native_chat` driver; only the provider-specific formats differ, via small per-kind seams. Shared machinery (`_build_system_prompt`, `_run_tool_calls`, `_maybe_lint_fix`) is defined once and reused by the Ollama loop too.
- Session persistence extracted from the `Kodiqa` God-class into a dedicated `SessionStore` (`session_store.py`); `Kodiqa` keeps thin wrappers, so behavior is unchanged. First step of splitting the God-class into focused units.
Tests
- Characterization tests for the chat-loop seams and `SessionStore` round-trips. 350 total; ruff + CI green.
kodiqa.py: 6251 → 5999 lines.
Weekly OSS security release digest.
The CVE patches and breaking changes that affected production tools this week. One email, every Sunday.
No spam, unsubscribe anytime.
Share this release
About Kodiqa Agent
All releases →Related context
Related tools
Beta — feedback welcome: [email protected]