Major Updates: Multi-Model Support & News Tool

Multi-model providers are integrated, supporting multiple models per API key across 4 capability-derived job groups: Vision, Reasoning, Analytical, and Utility.

Cognitive jobs can now be assigned models per job via a column in job_provider_assignments, and the Brain UI supports tag-style multi-model input and 4-group cognition views.

A new endpoint, PUT /providers/jobs/groups/<group_name>, allows for batch assignment of jobs to groups.

The news tool (Phase 1) is added, supporting 4 actions across 56 RSS feeds with category routing and deduplication.

A World Awareness Service (Phase 2) is introduced to run hourly, deriving user interests and writing signals to WorldStateService without LLM calls.

Refactoring occurred by deleting storage/retrieval facade shims and condensing the digest worker core logic.

Goal ecology improvements include using ThreadPoolExecutor for strategy generation, compacting outcome_feedback JSON, and using sentence-level embeddings for motive alignment.

Unit test coverage saw significant expansion across multiple files, including new tests for response parsing layers and chat history processing.

• Providers support multiple models (one API key, many model selections).

• 22 cognitive jobs are grouped into 4 capability groups.

• New news tool supports 4 actions across 56 public RSS feeds.

• World Awareness Service runs hourly to derive user interests.

• Refactored by deleting storage/retrieval facade shims.

• Goal ecology thread pool replaces unbounded threading.