ROADMAP.md

Roadmap

Deployment plan and milestones for prisma-chatbot. Living document — items move from Planned to In progress to Done as the project evolves.

Milestone 1 — Scaffolding

Done: Repo skeleton, docs, dependency manifest, env template, gitignore. No chatbot logic yet.

• Repo created, license added
• CLAUDE.md, README, ARCHITECTURE, ROADMAP drafts
• src/, tests/, assets/ directories with placeholder modules
• requirements.txt, .gitignore, .env.example

Milestone 2 — Minimal end-to-end loop

Done: Get a single message in / response + evaluation out working locally, even with a rough prompt.

• Implement src/config.py (model id, attributes, turn cap)
• Implement src/prompt.py (v1 dual-role prompt)
• Implement src/inference.py (HF Inference client wrapper)
• Implement src/evaluation.py (JSON parsing + validation)
• Implement app.py (minimal Gradio UI)
• First pytest tests for parsing/validation

Milestone 3 — Prompt and UX iteration

Done: Refine Prisma's voice, evaluation display, and the "check what I think" affordance.

• Refine system prompt for voice consistency and structured-output reliability
• Decide evaluation display: numeric, verbal, or both
• Decide update cadence: live each turn vs. on-request

Milestone 4 — Public deployment

Done: Shipped to a Hugging Face Space and linked from the README.

• Hugging Face Space configuration
• Per-session turn cap
• Public URL added to README and papers

Milestone 5 — Stretch ideas

Planned: Explicitly non-blocking; consider only after the demo is live and stable.

• About panel copy (assets/about.md)
• Rate limiting (per-IP / per-session beyond the turn cap)
• Light usage analytics (anonymous, aggregate)
• "Compare models" mode
• Downloadable conversation + evaluation transcript
• Linguistic feature highlighting (which words/choices shifted scores)

Future work (v2+)

Attribute customization

• User selects up to 6 attributes from a curated extended list (~15–20 dimensions covering social, cognitive, affective, and pragmatic perception)
• Extended list includes: pushy, knowledgeable, well-prepared, pedantic, helpful, arrogant, warm, evasive, confident, anxious, etc. (final set TBD)
• v1's default set remains as the "quick start" option