feat: add AI research assistant suite module

ai-research-assistant-suite/.env.example

@@ -0,0 +1,5 @@
+PORT=3002
+SCIBASE_AI_ASSISTANT_DEMO_MODE=true
+REVIEW_TEMPLATE_DOMAIN=biology
+REPRODUCIBILITY_SANDBOX=docker-disabled-demo
+GAP_FINDER_CORPUS_VERSION=demo-2026-05

ai-research-assistant-suite/.gitignore

@@ -0,0 +1,5 @@
+node_modules/
+dist/
+coverage/
+.env
+.DS_Store

ai-research-assistant-suite/README.md

@@ -0,0 +1,51 @@
+# SCIBASE.AI AI Research Assistant Suite
+
+Self-contained implementation for issue #16. The module provides deterministic assistant workflows that can later be backed by hosted LLMs or scientific models:
+
+- Auto peer review reports with domain templates, review-dataset profile metadata, statistical checks, ethics/data availability checks, citation checks, and claims-versus-evidence flags
+- Reproducibility confidence checks across notebooks, execution environments, pinned dependencies, dataset manifests, pipeline artifacts, validation splits, expected outputs, run commands, deterministic seeds, and previous attempts
+- Research gap finder using project keywords, citation signals, replication counts, unresolved questions, limitations, and negative-result follow-ups from a seed corpus
+- Combined project dashboard with review score, reproducibility score, top gap novelty, and readiness score
+- JSON APIs and demo UI for reviewer smoke tests
+- Unit tests for review generation, reproducibility scoring, sandbox metadata, gap discovery, and dashboard composition
+
+## Run locally
+
+```bash
+cd ai-research-assistant-suite
+npm install
+npm run dev
+```
+
+Open `http://localhost:3002`.
+
+## Verify
+
+```bash
+npm run typecheck
+npm test
+npm run build
+```
+
+## Demo API
+
+```bash
+curl -s http://localhost:3002/api/dashboard/project_organoid_response | jq .
+curl -s http://localhost:3002/api/review/project_organoid_response | jq .
+curl -s http://localhost:3002/api/reproducibility/project_organoid_response | jq .
+curl -s http://localhost:3002/api/gaps/project_organoid_response | jq .
+```
+
+## Demo artifact
+
+- `docs/demo/dashboard.png`
+- `docs/demo/ai-research-assistant-suite-demo.mp4`
+
+## Production integration notes
+
+The deterministic services are designed as stable boundaries for production model adapters:
+
+- `generatePeerReview` exposes the model/rubric profile and can accept LLM-generated findings after schema validation.
+- `runReproducibilityChecks` already returns sandbox metadata and previous-attempt links; production can replace the local simulator with a real notebook/container runner.
+- `findResearchGaps` can swap the seed corpus for vector search, citation graph traversal, or Semantic Scholar/PubMed connectors while preserving the same opportunity schema.
+- The current implementation avoids raw private-content export and uses project-local artifacts only.

ai-research-assistant-suite/docs/demo-script.md

@@ -0,0 +1,9 @@
+# Demo script
+
+1. Start with `npm run dev`.
+2. Open `http://localhost:3002`.
+3. Show the selected project and keyword context.
+4. Show the review score and findings table.
+5. Show reproducibility confidence and the recommended reproduction command.
+6. Show top research gaps ranked by novelty.
+7. Run `npm test` to verify deterministic assistant logic.