This is the discovery index for QuantMind's public operations, supported sources, examples, design documents, and verification commands. Source-specific acquisition mechanics remain internal unless they are intentionally documented as a public preprocessing primitive.
Public callable names follow the operation naming contract. The runtime API serves Python callers; coding-agent guidance lives in the repository development harness.
| Operation | Import | Input and config | Result | Example | Design or guide |
|---|---|---|---|---|---|
| Source-first paper flow | quantmind.flows.paper_flow |
PaperInput, PaperFlowCfg |
PaperFlowResult |
Persist and search a paper | Paper flow design |
| Paper structure build | quantmind.flows.PaperFlow |
PaperFlow(PaperStructureCfg); build(): PaperInput |
PaperStructureTree (self-contained) |
Build and retrieve | Structure retrieval design |
| Reasoning-based retrieval (agentic) | quantmind.mind.AgenticRetriever |
AgenticRetriever(RetrievalCfg); retrieve(): one StructureTree + question (no library) |
list[RetrievalEvidence] |
Build and retrieve | Structure retrieval design |
| News collection | quantmind.flows.collect_news |
NewsWindow, NewsCollectionCfg |
NewsBatch from quantmind.preprocess |
Collect news | News collection design |
| Bounded fan-out | quantmind.flows.batch_run |
Operation inputs and shared config | BatchResult |
README usage | API docstrings |
| Local semantic search | quantmind.library.LocalKnowledgeLibrary |
BaseKnowledge or PaperFlowResult, SemanticQuery |
list[SemanticHit] |
Library example | Library guide |
| Page-aware document RAG | quantmind.rag.chunk_parsed_document, quantmind.rag.retrieve_parsed_document |
ParsedDocument, splitter config, and query |
tuple[ParsedDocumentHit, ...] |
Paper RAG | Document RAG design |
Import public inputs and configs from quantmind.configs, flow operations and
builders from quantmind.flows, and cognitive services from quantmind.mind.
Import result contracts from the canonical layer shown in the catalog.
| Source | Source selection | Operation | Live-network component smoke test |
|---|---|---|---|
| PR Newswire | NewsWindow(source="pr-newswire", ...) |
collect_news |
python scripts/verify_news_e2e.py |
| arXiv Transformer PDF | ArxivIdentifier(id="1706.03762v7") |
paper_flow, persistence, reopen, search, and resolution |
python scripts/verify_pdf_rag_e2e.py |
| Golden paper PDF (structure) | LocalFilePath(...golden/paper.pdf) |
PaperFlow.build, standalone put/open_structure, and AgenticRetriever.retrieve |
python scripts/verify_structure_e2e.py |
The PR Newswire smoke test checks the public RSS feed, a complete preceding
24-hour listing window, and ticker-hint recall on a bounded sample of up to 25
article pages. The news job in .github/workflows/e2e.yml runs daily, on
manual dispatch, and only on pull requests that change its dependency paths.
It is not a required merge check, so external PR Newswire availability cannot
block unrelated changes.
The paper-flow job fetches exact arXiv revision 1706.03762v7, preserves at
least 15 pages, runs bounded gpt-4o-mini summarization, persists summary and
chunk projections with text-embedding-3-small, reopens the database, searches
both artifact kinds, and resolves every hit. It runs daily, manually, and on
pull requests that change its dependency paths, and it remains non-required
because arXiv and model providers are public-network dependencies.
The structure job builds a self-contained PaperStructureTree from the local
golden fixture PDF under the default model, dumps and reopens it through the
library unchanged, and runs a real AgenticRetriever traversal. It exercises the
one path offline tests mock away (a real structured-output draft call and a real
agentic loop). It runs daily, manually, and on pull requests that change its
dependency paths, and it remains non-required because model providers are a
public-network dependency.
When the repository OPENAI_API_KEY secret is unavailable, the job emits an
explicit skip notice instead of reporting an implementation failure; the
catalog command remains the direct way to run the same bounded slice locally.
Run the deterministic required verification for every change:
bash scripts/verify.shIt covers formatting, linting, typing, import boundaries, unit tests, and
coverage, and must remain network-free. The required .github/workflows/ci.yml
workflow runs this same harness after file-hygiene hooks. When a change affects
a public-network component, also run every applicable live-network smoke test
listed above; .github/workflows/e2e.yml owns those component jobs.
Use the quantmind-dev component workflow. A public operation is not complete
until its typed contract, package exports, offline tests, focused example,
design or guide, and catalog row agree. A public-network source additionally
needs mocked source tests plus a bounded live verifier and component job in
.github/workflows/e2e.yml.
Each live component owns one scripts/verify_<component>_e2e.py command and
one named job in the existing e2e.yml. Extend the workflow's precise PR path
filter for the component. When multiple live jobs exist, use GitHub-native
per-job change detection so only affected component jobs run. Add commands only
to the catalog above; root agent guidance stays component-neutral. Do not
create a workflow per component or a generic E2E runner, registry, or base
class.