academic-search

name	academic-search
description	Query the keyless academic and public-domain registries (OpenAlex, Crossref, Semantic Scholar, Project Gutenberg, Internet Archive, Stack Exchange) for a topic and get back normalized, copyright-safe reference metadata (title, link, provenance). Use when gather-citations or verify-claims needs primary-literature grounding instead of the LLM's training memory.
metadata	{"agent":"research-factcheck","source":"Project Atlas Agent Skills Manifest §4 (academic grounding)","layer":"L5-content-verify"}

academic-search

Retrieve source-linked references from the keyless scholarly / public-domain registries, governed by the provider-policies cache + rate limit on the NestJS side.

When to use

• gather-citations needs a factual spine from primary literature (dossier grounding).
• verify-claims needs to check a draft's dates/names/causal claims against real scholarship.
• A "Hidden Legend" needs the 19th-century books or modern papers that document them.

Sources (server/src/modules/search/open-source-search.service.ts + AcademicSearchService)

• OpenAlex (CC0), Crossref (open bibliographic metadata) — routed through the polite pool via the contact mailto.
• Semantic Scholar (public tier) — title/year/authors/DOI only (no abstracts stored).
• Project Gutenberg (Gutendex) and Internet Archive — public-domain primary texts.
• Stack Exchange (History site by default) — attributable question metadata (title/link/score), used to surface the human debates around a topic, not for trends.

All six are keyless. Retrieval is governed by AcademicSearchService: a policy-gated 30-day metadata cache + per-source per-minute burst guard (server/src/modules/search/provider-policies/), so bulk polling is cheap and polite automatically. Copyright-safe: only facts / reference metadata are returned and stored, never article or book bodies (spec §2.5).

Function signature (manifest contract)

def academic_search(query: str, source: str | None = None) -> list:
    """Return normalized academic references [{source, title, url, meta}] for a query.
    If `source` is given (openalex|crossref|semanticscholar|gutenberg|archive|stackexchange),
    query only that registry; otherwise fan out across all six."""

Inputs / Outputs

• Input: query (topic/subject/claim phrase); optional source for a targeted single-registry lookup.
• Output: list of {source, title, url, meta} — meta is a short provenance string (year · author, etc.).

Implementation

Fan-out (all registries): temporal-bridge → start_workflow("academicSearchWorkflow", "trend-signals", [{"query": query}]), then poll get_workflow_result(workflow_id).

Single registry: start_workflow("academicSearchSourceWorkflow", "trend-signals", [{"source": source, "query": query}]).

Runs the academicSearch / academicSearchSource Temporal activities (server/src/modules/temporal/activities/academic-search.activities.ts) on the trend-signals task queue, which go through the governed AcademicSearchService (cache + rate limit apply automatically). This is the same reach pattern fetch-signals uses for SearchService.

Backend dependency

• No API keys required (all six registries are keyless). Redis powers the cache + burst limiter; a Redis outage fails open (re-fetches), never blocks.
• CONTACT_EMAIL (optional) sets the polite-pool mailto for OpenAlex/Crossref.

Model

No model call — this is deterministic retrieval + normalization. Feed the returned references into gather-citations / verify-claims as grounding context.