 Jon Saad-FalconandClaude Opus 4.6
|
323d7ff032
|
Add TOML config system for eval suites, pillar-aligned config, and documentation
- Eval config: TOML-based suite configs defining models x benchmarks matrix,
loaded via --config flag. Includes load_eval_config(), expand_suite(),
7 config dataclasses, 3 example configs, and 61 new tests.
- Pillar-aligned config: generation params in IntelligenceConfig, nested
engine/learning configs, agent objective/system_prompt/context_from_memory,
structured learning sub-policies, TOML migration layer.
- Documentation: evaluations user guide, evals API reference, updated
mkdocs.yml navigation, updated architecture docs.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
|
2026-02-24 03:34:05 +00:00 |
|
|
Jon Saad-FalconandClaude Opus 4.6
|
8d538cd1b0
|
Add orchestrator training, channels, LiteLLM engine, and simplify learning taxonomy
Major changes across parallel sessions:
- Add orchestrator SFT & GRPO training subpackage (learning/orchestrator/)
with episode types, multi-objective reward, prompt registry, policy model,
RL environment, and registered learning policies
- Add structured THOUGHT/TOOL/INPUT/FINAL_ANSWER mode to OrchestratorAgent
- Add 15 channel backends (Discord, Slack, Telegram, Email, Webhook, IRC,
Matrix, Teams, WhatsApp, Signal, Mattermost, BlueBubbles, Feishu,
Google Chat, Webchat) with channel tools and config
- Add LiteLLM engine backend for unified LLM provider access
- Add RLM agent and REPL tool
- Remove ToolLearningPolicy — learning taxonomy now only targets
Intelligence (LM weights/routing) and Agents (logic/ICL/tool strategies)
- Rename SFTPolicy to SFTRouterPolicy (backward-compat alias kept)
- Remove OpenClaw agent infrastructure (openclaw*.py, openclaw_bridge.py)
- Fix async streaming tests (asyncio.run vs deprecated get_event_loop)
- Fix server channel route tests (pytest.importorskip for optional fastapi)
- Track uv.lock for reproducibility
- Update CLAUDE.md and docs to reflect all changes
1676 tests pass, 37 skipped.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
|
2026-02-23 18:32:32 +00:00 |
|
|
Jon Saad-FalconandClaude Opus 4.6
|
852259f18b
|
Restructure codebase into 5-pillar architecture with MCP tool management, composition layer, and structured learning
Phase 1: Move RoutingContext to core/types.py, add RouterPolicy and QueryAnalyzer ABCs to intelligence/_stubs.py
Phase 2: Move memory backends to tools/storage/, convert memory/ to backward-compat shims
Phase 3: Add MCPToolAdapter, storage MCP tools, upgrade MCP server to spec 2025-11-25
Phase 4: Add SystemBuilder + JarvisSystem composition layer (system.py)
Phase 5: Add InstrumentedEngine for opt-in telemetry, simplify all agents
Phase 6: Add LearningPolicy ABC taxonomy with SFTPolicy, AgentAdvisorPolicy, ICLUpdaterPolicy
Phase 7: Update config schema (ToolsConfig, MCPConfig, TracesConfig, per-pillar learning policies)
Also: update all docs, README (DSPy-inspired), CLAUDE.md, and add logo assets.
1391 tests pass, 32 skipped. Zero new lint errors. Full backward compatibility via shims.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
|
2026-02-22 05:57:13 +00:00 |
|
|
Jon Saad-FalconandClaude Opus 4.6
|
301e9cd2d4
|
Implement OpenJarvis v1.0 — all five pillars, SDK, benchmarks, Docker
Complete implementation across six development phases (v0.1 through v1.0):
- Core: Registry system, config, event bus, types (Phase 0)
- Intelligence + Inference: Model routing, Ollama/vLLM/llama.cpp/Cloud engines (Phase 1)
- Memory: SQLite/FAISS/ColBERT/BM25/Hybrid backends, document ingest, context injection (Phase 2)
- Agents: Simple/Orchestrator/Custom/OpenClaw agents, tool system (Phase 3)
- Learning: HeuristicRouter, reward functions, GRPO stub, telemetry aggregation (Phase 4)
- SDK: Jarvis class, OpenClaw protocol/transport, benchmarks, Docker deployment (Phase 5)
520 tests passing, 8 skipped (optional deps). Ruff lint clean.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
|
2026-02-17 00:52:48 +00:00 |
|