smouj/aicodestudio/aminer-daily-paper

aminer-daily-paper

Personalized academic paper recommendation.

When to activate

Any time the user asks for paper recommendations:

• Explicit command: /aminer-dp, /aminer-dp topics: RAG, multimodal agents
• Natural language: recommend me recent papers on multimodal agents, suggest some papers on tool-use, I work on RAG, give me a few papers

Input parsing (done by the model)

Before calling the script, extract from the user input:

At least one of topics / author_name / aminer_author_id should be provided.

Call strategy

Execution

The script reads .z-ai-config (JSON) following the z-ai-web-dev-sdk convention, searching in this order:

1. ./.z-ai-config (cwd)
2. ~/.z-ai-config (home)
3. /etc/.z-ai-config (system)

Required fields: baseUrl, apiKey, token (JWT for X-Token). Optional: chatId, userId.

python3 "{baseDir}/scripts/recommend.py" \
  [--topic "multimodal agents"] \
  [--topic "tool-use"] \
  [--author-name "Jie Tang"] \
  [--author-org "Tsinghua University"] \
  [--aminer-author-id "696259801cb939bc391d3a37"] \
  [--size 5] \
  [--language-sort zh]

The script POSTs to ${baseUrl}/functions/invoke with headers Authorization: Bearer ${apiKey}, X-Z-AI-From: Z, and X-Token: ${token} (plus optional X-Chat-Id / X-User-Id), and prints the paper list as Markdown.

Output handling

• On success, the script prints a Markdown paper list — forward it directly to the user.
• On non-zero exit, stderr contains the error — relay it concisely to the user.

Error handling

• API returns an error → relay the error; do not switch to another skill.
• No results → suggest the user broaden topics or adjust the query.