mirror of
https://github.com/MervinPraison/PraisonAI.git
synced 2026-11-07 12:47:55 -10:00
Mervin Praisonandpraisonai-bot
b2611e5d61
Re-implements the CAJAL / scientific-paper functionality that was reverted from the core SDK (PR #1611 / revert #1641) as proper examples per AGENTS.md: - examples/scientific_writing/scientific_writer.py Single-agent pattern: regular Agent + @tool functions, no subclass. Uses the CAJAL-4B HuggingFace model via litellm. - examples/scientific_writing/multi_agent_paper.py Multi-agent workflow: Literature Reviewer -> Methodologist -> Writer composed with AgentTeam + sequential Tasks (no custom Agent classes). - examples/scientific_writing/agents.yaml YAML parity for the multi-agent workflow (AGENTS.md §7 '3-way' rule). Why this location: - AGENTS.md §4.1: core SDK has no heavy domain implementations. - AGENTS.md §4.4: compose specialised Agents + tools, don't subclass Agent. - AGENTS.md §6.1: domain-specific behaviour belongs in @tool functions. - Wrapper / praisonai-tools were alternatives; examples/ keeps it discoverable and avoids bloating either package. Co-authored-by: praisonai-bot <[email protected]>