James 2.5

Introducing James v2.5.0: The Autonomous Node.

We’ve successfully completed Phase 3 of our roadmap, transforming the multi-agent R.A.I.N. Lab into a persistent, self-healing research operating system.

​When you deploy a background swarm to monitor new pre-prints, we monitor the swarm.

Our new Self-Healing Heartbeat Monitor detects logic stagnation and automatically restarts the process, ensuring your overnight research runs to completion.

​24/7 Monitoring: Continuous ArXiv parsing with our autonomous_research.sh service.
​
Ritual Learning: James now teaches itself. Successful logic frameworks are saved as repeatable "skills" in the Action-to-Text Episodic Memory Graph.
​
Trusted Rigor: Local execution. Local data. Epistemic failsafes are always active.
​Your local knowledge graph is waiting to be built.

What's Changed

• feat(ollama): add --raw-command supervisor mode and autonomous research launcher by @topherchris420 in #145

• docs(onboarding): simplify entry point and clarify non-technical onboarding by @topherchris420 in #103
• fix(rain-lab): restore chat as default launcher mode by @topherchris420 in #105
• Add libp2p P2P module and publish agent advisory results by @topherchris420 in #107
• Introduce strict 5-stage research protocol and refactor agents/workflow by @topherchris420 in #108
• Add multilingual documentation hubs and comprehensive docs tree by @topherchris420 in #109
• Enforce estop before gateway/daemon start and set default Body daemon port to 4200 by @topherchris420 in #111
• Add ZeroClaw runtime bridge docs, Rust daemon client in orchestrator, and estop startup guard by @topherchris420 in #112
• Add Rust daemon bridge and agent registry support; enforce estop before starting runtime by @topherchris420 in #113
• fix: use non-login shell in service tests to avoid profile noise by @topherchris420 in #116
• feat(providers): add 17 providers + critical bug fixes from upstream by @topherchris420 in #118
• Remove Azure OpenAI provider and fix shell test flags by @topherchris420 in #119
• refactor: burn down 5 crate-level clippy suppressions by @topherchris420 in #120
• Make dev the primary integration branch; update CI, docs, and crate metadata by @topherchris420 in #121
• feat(tool): add ArXiv paper search tool for academic research by @topherchris420 in #122
• Add resonance/cymatic visualization for frequency discussions by @topherchris420 in #123
• Add episodic memory hook for recording agent tool calls by @topherchris420 in #125
• refactor: format shell command logging in episodic memory hook by @topherchris420 in #126
• Add adversarial peer-review swarm orchestrator for document critique by @topherchris420 in #127
• Add hypothesis tree, stagnation monitor, and MCP server for R.A.I.N. Lab by @topherchris420 in #128
• Fix test isolation in config_save_and_load_tmpdir by using unique UUIDs by @topherchris420 in #129
• Remove unused imports from mcp_server, tests, and stagnation_monitor by @topherchris420 in #131
• chore: bump version to 0.3.2 by @topherchris420 in #132
• docs: Add James 2 (v0.3.2) release notes by @topherchris420 in #133
• Add CI/CD workflow status badges to all README locales by @topherchris420 in #134
• Replace unwrap() calls with proper error handling throughout codebase by @topherchris420 in #135
• Revert "Replace unwrap() calls with proper error handling throughout codebase" by @topherchris420 in #136
• Codex-generated pull request by @topherchris420 in #137
• Improve lab orchestrator (scoring/reporting), add chat temp/max-tokens flags, update docs and tests by @topherchris420 in #140
• ci(workflows): run push checks on all branches by @topherchris420 in #141

Full Changelog: v1.1.0...v2.5