Pesquisa e Web

Structured entity data (people, places, concepts) from Wikidata knowledge graph.

Open-source deep research for AI agents: 40 channels, 10+ Chinese sources.

Search scholarly works through OpenAlex's public works search API with open-access URL fallback.

Add items (research subjects) to existing research outline.

Define the execution contract for isolating a research sub-task — input schema, budget, return summary, evidence list, failure status. Complements decompose-task (which only splits the problem) by giving each split a bounded, auditable execution unit the runtime AI can farm out to a sub-agent or parallel session.

Authoritative encyclopedia articles via the Wikipedia Action API (English edition).

Use for biomedical and life-science literature searches when query expects PubMed articles, clinical trials, or medical research papers.

Session-level policy for keeping the runtime AI's context window healthy across long research — keep-last-k tool results, offload older evidence to disk, trigger compaction at thresholds. Borrows MiroThinker's keep_tool_result, deepagents' summarization middleware, and deer-flow's SummarizationEvent pattern. Orthogonal to assemble-context (which is per-synthesis-step); this is per-session.

Open-source deep research for AI agents: 40 channels, 10+ Chinese sources.

Use to scrape a URL into clean Markdown when the page is JS-rendered, behind anti-bot, or a PDF — and FIRECRAWL_API_KEY is available. Degrades gracefully to warn when key is missing.

Open-source deep research for AI agents: 40 channels, 10+ Chinese sources.

Distill reusable knowledge from successful session tool-call traces. Reads runtime tool-call logs + Evidence outputs + user acceptance signals, filters successful paths, and writes compact patterns to per-skill experience/patterns.jsonl. Differs from outcome-tracker (which only records downstream acceptance) by parsing the full trace structure — inputs, tool choices, branches, recoveries.