Setup¶

This guide covers both full production-like setup and low-cost local development.

Prerequisites¶

Tool	Version	Install
Python	3.12+	python.org
uv	latest	`brew install uv`
Git	latest	git-scm.com

1. Clone the Repository¶

git clone https://github.com/chitrank2050/askchitrank.git
cd askchitrank

2. Create the Environment¶

make init
make install
cp .env.example .env.dev

3. Choose a Local Mode¶

Option A — Token-free local development¶

If you want to work on the API, prompt flow, streaming, or frontend integration without spending Groq or Voyage quota:

API_TOKEN=dev-token
DEV_MODE=true

Optional:

CONTACT_EMAIL=hello@example.dev

In this mode:

Groq calls are skipped
Voyage calls are skipped
chat can run without a database
fictional seeded data is used for local responses

This is the recommended default for day-to-day local iteration.

Option C — Local embedding fallback¶

If you want to use real local embeddings instead of Voyage AI (e.g., to save quota or work offline):

Install extra dependencies:
```
uv sync --extra local-embed
```

Update your .env.dev:

EMBEDDING_PROVIDER=local
EMBEDDING_DIMENSIONS=384

In this mode, sentence-transformers (all-MiniLM-L6-v2) handles all embedding tasks locally on your machine.

Option B — Full provider-backed development¶

If you want to exercise the full production-style pipeline, fill in:

GROQ_API_KEY=your-groq-api-key
VOYAGE_API_KEY=your-voyage-api-key
DATABASE_URL=postgresql+asyncpg://postgres:[PASSWORD]@db.[REF].supabase.co:5432/postgres
DATABASE_URL_SYNC=postgresql+psycopg2://postgres:[PASSWORD]@db.[REF].supabase.co:5432/postgres
SUPABASE_URL=https://[REF].supabase.co
SUPABASE_KEY=your-supabase-anon-key
SANITY_PROJECT_ID=your-project-id
SANITY_API_TOKEN=your-api-token
API_TOKEN=your-generated-token
DEV_MODE=false

4. Database Setup¶

For the full pipeline, follow Database Setup and run:

make db-migrate

If you stay in token-free DEV_MODE, chat-only API work does not require the database.

5. Prepare Data Sources¶

Resume¶

Place your resume PDF at data/resume.pdf.

LinkedIn¶

Export your LinkedIn data and place these CSVs in data/linkedin/:

Profile.csv
Recommendations_Received.csv

Sanity CMS¶

No manual content export is needed. Data is fetched via the Sanity API during ingestion.

6. Ingest Data¶

make ingest

Notes:

In full mode, this stores real embeddings in the database.
In DEV_MODE, ingestion uses fictional seeded source content.
In DEV_MODE without a configured database, ingestion is skipped because there is nowhere to persist chunks.

7. Start the API¶

make api

If DEV_MODE=true, the API can start and serve fictional seeded chat responses even without database credentials.

Service Setup¶

Service	URL	Cost
Groq	console.groq.com	free tier
Voyage AI	voyageai.com	generous free tier
Supabase	supabase.com	free tier
Sanity	sanity.io/manage	depends on workspace plan

Developed by Chitrank Agnihotri