martin65536/liquid-glass-webgl/aminer-daily-paper

aminer-daily-paper

Personalized paper recommendation via AMiner rec5 API. Token required: set AMINER_API_KEY env var.

• Docs: https://open.aminer.cn/open/docs | Console: https://open.aminer.cn/open/board?tab=control

When to activate: any time the user asks for paper recommendations — explicit command (/aminer-dp ...) or natural language (recommend me papers on RAG, 帮我推荐最近的多模态论文).

Pre-flight: Check Required Environment Variables

`AMINER_API_KEY` — Always required. Check before calling the script:

[ -z "${AMINER_API_KEY+x}" ] && echo "AMINER_API_KEY missing" || echo "AMINER_API_KEY exists"

If missing, stop and tell the user:

AMINER_API_KEY is not set. Please obtain a token at https://open.aminer.cn and set it as an environment variable.

No other environment variables are required.

API Endpoint

POST https://publicapi.chatglm.cn/chatglm_public/skill/aminer/api/v3/paper/rec5
Authorization: ${AMINER_API_KEY}
Content-Type: application/json;charset=utf-8

Request Fields

At least one of author_name or topics should be provided. When none are given, the API returns personalized recommendations based on the account associated with AMINER_API_KEY.

Response Structure

{
  "code": 200,
  "success": true,
  "data": [{
    "offset": 0,
    "size": 5,
    "total": 32,
    "papers": [{
      "paper_id": "...",
      "arxiv_id": "",
      "title": "...",
      "year": 2026,
      "authors": ["Author A", "Author B"],
      "keywords": ["kw1", "kw2"],
      "summary": "...",
      "structured_summary": {
        "research_problem": "...",
        "research_challenge": "...",
        "research_method": "...",
        "experimental_results": ""
      },
      "famous_authors": [],
      "aminer_author_profiles": [],
      "author_entries": [],
      "links": {
        "aminer": "https://www.aminer.cn/pub/{paper_id}",
        "arxiv": "",
        "pdf": ""
      },
      "paper_url": "https://www.aminer.cn/pub/{paper_id}",
      "source": "local_rec5"
    }]
  }]
}

Input Formats

Structured commands or plain natural language — both are valid.

/aminer-dp
/aminer-dp topics: multimodal agents, tool-use
/aminer-dp scholar: Jie Tang org: Tsinghua papers: OAG-Bench | RPC-Bench
recommend me recent papers on RAG

/aminer-dp with no parameters calls the API with only the token — the API uses AMINER_API_KEY to identify the account and returns personalized recommendations.

Natural language input — you (the model) must parse it into fields before calling the script. Critical for `topics`:

1. `topics` — do not “translate away” the user’s intent

• If the user already wrote topics: in the trigger (e.g. 具身智能, 环境保护), pass those exact strings into handle_trigger.py’s --text. Do not replace them with unrelated English terms (e.g. do not map arbitrary topics to “Knowledge Distillation”, “Smart agriculture”, or any other field the user did not ask for).
• If you add English for retrieval, it must be a faithful alias of the same concept (e.g. 具身智能 → embodied intelligence, 环境保护 → environmental protection). When in doubt, keep the user’s original words and do not invent synonyms.
• Never change the user’s topic into a different research area.

1. Scholars and institutions (person search still English-oriented)

• author_name / author_org: use commonly used English forms when resolving scholars (e.g. Jie Tang, Tsinghua University), expand well-known institution abbreviations to full official names, and add author_org when the name is ambiguous. If you cannot map a name safely, ask the user.

1. `language_sort` — Put language_sort: zh or language_sort: en in the trigger only if the user clearly wants recommendations ranked with a Chinese or English preference. If they did not ask, do not add it (the API call omits language_sort).

1. Decide size and whether to make multiple calls (see Call Strategy).
2. Reconstruct the trigger, then call handle_trigger.py.

Example (Chinese topics — keep as-is):

• User: /aminer-dp topics: 具身智能, 环境保护
• You call: handle_trigger.py --text "/aminer-dp topics: 具身智能, 环境保护"

(Do not rewrite topics into unrelated English.)

Example:

• User: /aminer-dp 我做多模态智能体和 tool-use，帮我推荐最近论文
• You extract: topics: multimodal agents, tool-use
• You call: handle_trigger.py --text "/aminer-dp topics: multimodal agents, tool-use size: 5"

Example (scholar):

• User: /aminer-dp 我是唐杰，清华大学，做多模态和知识图谱
• You extract: scholar: Jie Tang, org: Tsinghua University, topics: multimodal, knowledge graph
• You call: handle_trigger.py --text "/aminer-dp scholar: Jie Tang org: Tsinghua University topics: multimodal, knowledge graph"

Example (ambiguous name, ask user):

• User: /aminer-dp 推荐张伟方向的论文
• You: "张伟是一个常见名字，请提供机构信息以便精确匹配，例如：张伟，北京大学。或者直接提供 aminer_author_id。"

`papers` field: representative paper titles (e.g. papers: OAG-Bench | RPC-Bench) accompany scholar/author_name for disambiguation context. They do not map directly to an API field.

Call Strategy

You decide size and whether to make multiple calls based on the input:

Multi-call rules:

• Call handle_trigger.py once per topic group, passing a focused topics: subset each time.
• Keep each topics: list to 1–3 closely related terms for precision.
• Make calls sequentially; present all results together after all calls finish.
• Total papers across all calls should not exceed ~15 unless the user asks for more.

Execution

Only one supported entrypoint:

python3 "{baseDir}/scripts/handle_trigger.py" \
  --base-dir "{baseDir}" \
  --text "<trigger text with explicit fields>" \
  [--config /path/to/config.yaml]

• --text: reconstructed trigger with explicit fields (topics:, scholar:, etc.)
• --config: optional path to a YAML config (defaults to {baseDir}/config.yaml when the file exists, via the runtime copy under outputs/)

handle_trigger.py parses the fields, calls the rec5 API, and returns JSON including reply_text (Markdown) for you to show to the user.

Contract

• Every explicit invocation is a new run.
• Do not answer with status-only text.
• Do not search, install, or repair skills.
• After running handle_trigger.py, check final_response in the JSON output:
• TEXT — Normal path. Present reply_text (Markdown) to the user. Optional: you may still refine wording for the active channel; prompts/enrich.md is a reference for Chinese enrichment if you want richer copy.
• Any error → report the reply_text (or error detail) to the user.

Note: The skill only returns JSON with reply_text; it does not implement channel-specific sending.

Error Handling

• AMINER_API_KEY missing → stop, prompt user to set it.
• No profile input → prompt user to provide topics, scholar name, or aminer_author_id.
• API error → report the error stage; do not fall back to other skills.