Overview & Products Page Updates - Graph-Aware Focus

Summary

Restructured the documentation to focus on graph-aware embeddings as the core value proposition, with clear documentation of both integration approaches (full integration vs. plugin).

Changes Made

✅ Logo now links to https://papr.ai/landing
✅ Added "Products" link to navbar
✅ Added "Dashboard Login" link to navbar
✅ Removed pricing page

2. Overview Page (overview/index.md)

Structure: Clean, card-based layout similar to Stripe

Common use cases (8 cards):

Document Q&A
Add graph-aware search to your stack (moved up, strengthened messaging)
Conversational AI with memory
Code search
Fraud detection (NEW)
Recommendation systems (NEW)
Knowledge management for teams
Scientific claim verification

Products (simplified to 4 main products):

Graph-Aware Vector Search
Knowledge Graphs
Document Intelligence
Chat Memory & Compression

Other sections: Start building (4 cards), Deployment options (3 cards), More resources (6 cards)

Removed: "How it works" section (moved to guides)

3. Products Page (overview/products.md)

Completely restructured to focus on 4 main products with clear documentation of integration approaches.

Graph-Aware Vector Search

Two Integration Approaches Documented:

Approach 1: Papr Storage (Full Integration)

Use when: Building new apps, want full memory management
Endpoints: POST /v1/memory with enable_holographic=true
Example: Add/search with graph-aware embeddings enabled

# Add
papr.memory.add(
    content="...",
    enable_holographic=True,
    frequency_schema_id="cosqa"
)

# Search
results = papr.memory.search(
    query="...",
    holographic_config={
        "enabled": True,
        "frequency_schema_id": "cosqa",
        "frequency_filters": {"language": 0.9}
    }
)

Approach 2: Plugin/Transform (BYOE)

Use when: Existing embeddings (OpenAI, Cohere), don't want to migrate
Endpoints:
- POST /v1/holographic/transform - Transform embeddings
- POST /v1/holographic/rerank - Rerank search results
Two options documented:
- Option A: Transform embeddings before storing in your vector DB
- Option B: Rerank existing search results

# Transform
transformed = papr.holographic.transform(
    embedding=openai_embedding,
    frequency_schema_id="cosqa",
    content="..."
)

# Rerank
reranked = papr.holographic.rerank(
    query="...",
    results=your_results,
    frequency_schema_id="cosqa"
)

Knowledge Graphs

Focus on fraud detection, recommendations, relationship mapping
Clear "When to use" and key capabilities

Document Intelligence

Three integration options: standalone, + vector search, + knowledge graphs
Pluggable extraction providers

Chat Memory & Compression

Session-based storage with automatic compression
Works with vector search for cross-session recall

4. Use Case Examples Updated

Code search example now shows both approaches:

Approach 1: Using Papr storage with enable_holographic=true
Approach 2: Plugin with existing embeddings using /v1/holographic/transform

Key Messaging

What Changed

❌ Before: "7 products" messaging (feature counting)
✅ After: Clear focus on 4 main products with graph-aware as differentiation

Integration Clarity

✅ Two clear paths: Full integration OR plugin approach
✅ When to use each: Documented decision criteria
✅ All endpoints documented: Both /v1/memory and /v1/holographic approaches
✅ Code examples: Show both paths side-by-side

Simplification

Removed complex product matrices and dependency diagrams
Removed "How it works" from overview (lives in guides)
Focused on action-oriented content (what you can build)
Card-based layout for scannability

Validation

✅ Dev server running at http://127.0.0.1:4000
✅ 0 markdoc errors
✅ All changes hot-reloaded and visible
✅ 467 total pages
⚠️ 83 pre-existing broken links (not introduced by these changes)

Next Steps (If Approved)

Review in browser at http://127.0.0.1:4000
Commit changes with descriptive message
Consider updating graph-aware-embeddings.md guide to match this structure
Update decision tree to reflect new product focus