Initial commit: multimodal RAG guide with Claude Code
Prompt-driven guide for building multimodal search using Gemini Embedding 2 + Pinecone + Claude Code. Includes example data (NASA public domain), step-by-step prompts, concepts explainer, cost breakdown, and troubleshooting guide.
This commit is contained in:
commit
edcd1721df
19 changed files with 4446 additions and 0 deletions
67
costs.md
Normal file
67
costs.md
Normal file
|
|
@ -0,0 +1,67 @@
|
|||
# What Does This Cost?
|
||||
|
||||
Short answer: you can do this entire guide for free or near-free.
|
||||
|
||||
## Claude Code
|
||||
|
||||
You need a Claude subscription that includes Claude Code.
|
||||
|
||||
- **Claude Pro ($20/month):** Includes Claude Code with usage limits.
|
||||
- **Claude Max ($100/month or $200/month):** Higher limits.
|
||||
|
||||
If you already have a Claude subscription, there is no extra cost.
|
||||
Claude Code handles building the app AND answering questions
|
||||
based on your search results.
|
||||
|
||||
## Google Gemini Embedding 2 (for embeddings only)
|
||||
|
||||
We use Google's Gemini Embedding 2 to convert your content into
|
||||
searchable embeddings. We do NOT use Gemini as a language model.
|
||||
Claude does the thinking. Gemini just creates the fingerprints.
|
||||
|
||||
- **Free tier:** Available through Google AI Studio with rate limits
|
||||
(1,500 requests per day)
|
||||
- **Paid:** $0.20 per million tokens
|
||||
|
||||
For this guide with 7 example files, you will use roughly 10-20
|
||||
requests total. The free tier is more than enough.
|
||||
|
||||
Why Gemini Embedding 2 and not Voyage (Anthropic's embeddings)?
|
||||
Because Gemini Embedding 2 supports text, images, AND video natively.
|
||||
Voyage supports text and images but not video. For a multimodal guide,
|
||||
the broadest format support wins.
|
||||
|
||||
## Pinecone Vector Database (free tier)
|
||||
|
||||
Pinecone's free tier includes:
|
||||
|
||||
- **Free:** 1 project, 5 indexes, 100,000 vectors
|
||||
- **No credit card required**
|
||||
|
||||
For this guide, we store about 15-30 vectors (one per chunk of
|
||||
content). You could store thousands of documents and still stay
|
||||
within the free tier.
|
||||
|
||||
## Total cost for this guide
|
||||
|
||||
| Service | What it does | Cost |
|
||||
|---------|-------------|------|
|
||||
| Claude Code | Builds the app, answers questions | Part of your subscription |
|
||||
| Gemini Embedding 2 | Converts content to searchable vectors | Free (well within free tier) |
|
||||
| Pinecone | Stores and searches vectors | Free (well within free tier) |
|
||||
| **Total** | | **$0 extra** |
|
||||
|
||||
## What if you want to scale up?
|
||||
|
||||
If you move beyond the example and want to index thousands of
|
||||
documents, here is what the costs look like:
|
||||
|
||||
| Scale | Gemini embeddings | Pinecone | Monthly total |
|
||||
|-------|-------------------|----------|---------------|
|
||||
| 100 documents | Free | Free | $0 |
|
||||
| 1,000 documents | ~$0.50 | Free | ~$0.50 |
|
||||
| 10,000 documents | ~$5 | Free | ~$5 |
|
||||
| 100,000 documents | ~$50 | $70+ (Starter plan) | ~$120 |
|
||||
|
||||
For most personal and small business use cases, you will stay
|
||||
comfortably in the free tier.
|
||||
Loading…
Add table
Add a link
Reference in a new issue