Which documents go into which bot

Paint-by-numbers for tomorrow’s AnythingLLM build. For each bot: the workspace name, the system prompt to paste, and the knowledge documents to load. Then create the embed and paste it into the bot’s index.qmd.

Upload method. AnythingLLM ingests .md / .html / .txt natively, so the simplest path is to upload these files directly in the AnythingLLM UI (drag-drop into the workspace, then “Save & Embed”). botstash is great for LMS exports but its scanner skips .md/.html (only pdf/docx/pptx/vtt/qti/imscc), so to use botstash you’d first render these to PDF. For Friday, UI upload is the reliable route.

System prompt = the contents of that bot’s bots/<name>/prompt.txt plus the DELIVERY MODE line from ANYTHINGLLM_SETUP.md. botstash does not set this; paste it in the workspace’s chat settings.

Every bot also gets these two shared docs: _backstories/retailflow_company_overview.md and _backstories/delivery_scenario_context.md.

Do NOT upload the red-herring / trap docs to any bot (the newsletter, the 2024 offsite notes, the 2023 pricing memo, the chatty emails). Those exist for the human document-discernment exercise; giving them to a bot would make it cite stale or irrelevant info. Bots get the accurate docs only; they already “know” the underlying facts from their profile docs.

Priority for Friday

Stand up these five (the afternoon stakeholders): Priya, Marcus, Emma, Tom, David. Sarah and Lisa are optional. One bot (Priya or Tom) is enough for the morning demo.

---

Priya Sharma, Head of Data & Analytics bots/priya_sharma/

The data-readiness oracle every group interviews in Sprint 1, about their initiative, so she gets all four data profiles. - _backstories/priya_sharma_data_analytics.md - documents/chatbot/cs-data-profile.html - documents/pricing/pricing-data-profile.html - documents/inventory/inventory-data-profile.html - documents/fraud/fraud-data-profile.html - documents/company/it-systems-overview.html

Marcus Kim, CIO bots/marcus_kim/

Tech ambition, legacy systems, vendor/build view across initiatives. - _backstories/marcus_kim_cio.md - documents/company/it-systems-overview.html - documents/inventory/legacy-systems-note.html - All four board memos: documents/{chatbot,pricing,inventory,fraud}/board-memo-*.html - Vendor proposals: documents/{chatbot,pricing,fraud}/vendor-proposal-*.html

Emma Rodriguez, Managing Director bots/emma_rodriguez/

Board promise, scope, what was committed. - _backstories/emma_rodriguez_ceo.md - All four board memos: documents/{chatbot,pricing,inventory,fraud}/board-memo-*.html - documents/company/company-fact-sheet.html - documents/company/ai-principles-draft.html

David Chen, CFO bots/david_chen/

Budgets, costs, ROI, go/no-go criteria. - _backstories/david_chen_cfo.md - All four board memos: documents/{chatbot,pricing,inventory,fraud}/board-memo-*.html - Vendor proposals (for costs): documents/{chatbot,pricing,fraud}/vendor-proposal-*.html - documents/company/ai-principles-draft.html

Tom Walsh, Customer Service Manager bots/tom_walsh/

Frontline reality; chatbot-focused (Sprint 2 inserts land on his team). - _backstories/tom_walsh_cs_manager.md - documents/chatbot/board-memo-chatbot.html - documents/chatbot/cs-data-profile.html - documents/chatbot/returns-policy.html - documents/company/org-chart.html

Sarah Thompson, COO bots/sarah_thompson/ (optional)

Operational reality, especially inventory/store ops. - _backstories/sarah_thompson_coo.md - documents/inventory/board-memo-inventory.html - documents/inventory/stock-replenishment-policy.html - documents/company/company-fact-sheet.html

Lisa Nguyen, GM Customer Experience bots/lisa_nguyen/ (optional)

Customer experience lens. - _backstories/lisa_nguyen_cco.md - documents/chatbot/returns-policy.html - documents/company/company-fact-sheet.html

---

After uploading each bot

1. Set the system prompt (prompt.txt + DELIVERY MODE line).
2. In AnythingLLM, enable an embeddable chat widget for the workspace and copy the <script> snippet.
3. Paste it into bots/<name>/index.qmd, replacing the <!-- AnythingLLM embed will go here --> placeholder block.
4. quarto render (the bot pages are Quarto-rendered) and deploy.

Paths above are relative to the sites/retailflow/ repo root.