scitrail 0.0.1

Analyze scientific literature with structured criteria to reveal the current state of research.

scitrail

scitrail builds an easy-to-read markdown report of the top voices in a field, including ORCID identifiers and a concise state-of-the-art executive summary.

The workflow is:

1. Read a YAML config (institution, optional department/departments, topic/topics, and simple knobs).
2. Resolve the institution in OpenAlex (with ROR when available).
3. Scan topic-relevant literature in that institution.
4. Rank the top voices (configurable count).
5. Summarize each voice and produce an executive summary.
6. Render a markdown report.

Install

uv sync

Optional local model support (for local LLM summarization with instructor):

uv sync --extra local-llm

Example config

institution: CU Anschutz
departments:
  - Department of Biomedical Informatics
topics:
  - Quantum
  - Artificial intelligence
max_people: 5
works_per_person: 8
lookback_years: 5
openalex_email: [e-mail]
# openalex_api_key: <optional>
llm:
  enabled: false

CLI

Generate a markdown report file:

uv run scitrail generate --config examples/cu_quantum.yaml --output report.md

Preview markdown in terminal:

uv run scitrail preview --config examples/cu_quantum.yaml

Compare department-scoped vs all-departments runs:

uv run scitrail generate --config examples/cu_topics_dbmi.yaml --output examples/cu_topics_dbmi_report.md
uv run scitrail generate --config examples/cu_topics_all_departments.yaml --output examples/cu_topics_all_departments_report.md

Run the built-in example end-to-end and write examples/cu_quantum_report.md:

uv run scitrail example

Execute the docs notebook version (renders rich markdown output in the notebook):

uv run --group docs --group notebooks poe docs-notebook-cu-quantum

Notes

• OpenAlex API keys are recommended for non-trivial usage.
• If local LLM dependencies are unavailable, scitrail automatically falls back to deterministic summaries so reports still complete.