scriptoriumadmin/supermemory
1
0
Fork 0
mirror of https://github.com/supermemoryai/supermemory.git synced 2026-09-08 22:21:07 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

Delete apps/docs/search/examples/hybrid-search.mdx

Browse source 
Co-Authored-By: mintlify[bot] <109931778+mintlify[bot]@users.noreply.github.com>
This commit is contained in:
mintlify[bot] 2025-12-19 00:31:41 +00:00 • committed by GitHub
parent 856e06dfc8
commit 3af34c2cee
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
1 changed files with 0 additions and 487 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

487
apps/docs/search/examples/hybrid-search.mdx
View file

@ -1,487 +0,0 @@
---
title: Hybrid search examples
description: Examples of using hybrid search mode to search memories and document chunks
---
Hybrid search mode searches memories first, then falls back to document chunks when needed. This provides comprehensive results from both structured memories and raw document content.
## Basic hybrid search
Search for information across both memories and document chunks:
<CodeGroup>
```bash cURL
curl -X POST https://api.supermemory.ai/v4/search \
-H "Authorization: Bearer YOUR_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"q": "What are the key features of machine learning?",
"searchMode": "hybrid",
"limit": 10,
"threshold": 0.7
}'
```
```javascript JavaScript
const response = await fetch('https://api.supermemory.ai/v4/search', {
method: 'POST',
headers: {
'Authorization': 'Bearer YOUR_API_KEY',
'Content-Type': 'application/json'
},
body: JSON.stringify({
q: 'What are the key features of machine learning?',
searchMode: 'hybrid',
limit: 10,
threshold: 0.7
})
});
const data = await response.json();
console.log(data);
```
```python Python
import requests
response = requests.post(
'https://api.supermemory.ai/v4/search',
headers={
'Authorization': 'Bearer YOUR_API_KEY',
'Content-Type': 'application/json'
},
json={
'q': 'What are the key features of machine learning?',
'searchMode': 'hybrid',
'limit': 10,
'threshold': 0.7
}
)
data = response.json()
print(data)
```
</CodeGroup>
### Response
```json
{
"results": [
{
"id": "mem_abc123",
"memory": "Machine learning key features: pattern recognition, automated learning, predictive modeling",
"metadata": {
"category": "ml_concepts",
"source": "training_notes"
},
"updatedAt": "2024-01-15T10:30:00Z",
"version": 1,
"rootMemoryId": null,
"similarity": 0.94,
"context": {
"parents": [],
"children": []
},
"documents": [],
"chunks": []
},
{
"id": "chunk_xyz789",
"chunk": "Machine learning algorithms can identify patterns in data without being explicitly programmed. Key features include supervised learning, unsupervised learning, and reinforcement learning approaches.",
"metadata": {
"source": "ml_textbook.pdf",
"page": 12
},
"updatedAt": "2024-01-14T09:00:00Z",
"similarity": 0.89,
"version": 1,
"context": {
"parents": [],
"children": []
},
"documents": [
{
"id": "doc_123",
"title": "Introduction to Machine Learning",
"type": "pdf",
"createdAt": "2024-01-10T08:00:00Z",
"updatedAt": "2024-01-14T09:00:00Z"
}
],
"chunks": []
}
],
"timing": 245,
"total": 2
}
```
## Hybrid search with document metadata
Include document information to understand the source of chunk results:
<CodeGroup>
```bash cURL
curl -X POST https://api.supermemory.ai/v4/search \
-H "Authorization: Bearer YOUR_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"q": "neural network architectures",
"searchMode": "hybrid",
"limit": 5,
"threshold": 0.75,
"include": {
"documents": true,
"summaries": true
}
}'
```
```javascript JavaScript
const response = await fetch('https://api.supermemory.ai/v4/search', {
method: 'POST',
headers: {
'Authorization': 'Bearer YOUR_API_KEY',
'Content-Type': 'application/json'
},
body: JSON.stringify({
q: 'neural network architectures',
searchMode: 'hybrid',
limit: 5,
threshold: 0.75,
include: {
documents: true,
summaries: true
}
})
});
const data = await response.json();
```
```python Python
import requests
response = requests.post(
'https://api.supermemory.ai/v4/search',
headers={
'Authorization': 'Bearer YOUR_API_KEY',
'Content-Type': 'application/json'
},
json={
'q': 'neural network architectures',
'searchMode': 'hybrid',
'limit': 5,
'threshold': 0.75,
'include': {
'documents': True,
'summaries': True
}
}
)
data = response.json()
```
</CodeGroup>
### Response
```json
{
"results": [
{
"id": "chunk_def456",
"chunk": "Common neural network architectures include feedforward networks, convolutional neural networks (CNNs), and recurrent neural networks (RNNs). Each architecture is optimized for different types of tasks.",
"metadata": {
"source": "deep_learning_guide.pdf",
"chapter": "3"
},
"updatedAt": "2024-01-16T14:20:00Z",
"similarity": 0.91,
"version": 1,
"context": {
"parents": [],
"children": []
},
"documents": [
{
"id": "doc_456",
"title": "Deep Learning Architectures",
"type": "pdf",
"metadata": {
"author": "Dr. Jane Smith",
"year": 2024
},
"summary": "A comprehensive guide to modern deep learning architectures, covering CNNs, RNNs, transformers, and more.",
"createdAt": "2024-01-15T10:00:00Z",
"updatedAt": "2024-01-16T14:20:00Z"
}
],
"chunks": []
}
],
"timing": 198,
"total": 1
}
```
## Hybrid search with filtering
Filter results by metadata while using hybrid search:
<CodeGroup>
```bash cURL
curl -X POST https://api.supermemory.ai/v4/search \
-H "Authorization: Bearer YOUR_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"q": "optimization techniques",
"searchMode": "hybrid",
"limit": 10,
"threshold": 0.7,
"filters": {
"and": [
{
"key": "category",
"operator": "equals",
"value": "deep_learning"
},
{
"key": "difficulty",
"operator": "less_than_or_equal",
"value": 3
}
]
}
}'
```
```javascript JavaScript
const response = await fetch('https://api.supermemory.ai/v4/search', {
method: 'POST',
headers: {
'Authorization': 'Bearer YOUR_API_KEY',
'Content-Type': 'application/json'
},
body: JSON.stringify({
q: 'optimization techniques',
searchMode: 'hybrid',
limit: 10,
threshold: 0.7,
filters: {
and: [
{
key: 'category',
operator: 'equals',
value: 'deep_learning'
},
{
key: 'difficulty',
operator: 'less_than_or_equal',
value: 3
}
]
}
})
});
const data = await response.json();
```
```python Python
import requests
response = requests.post(
'https://api.supermemory.ai/v4/search',
headers={
'Authorization': 'Bearer YOUR_API_KEY',
'Content-Type': 'application/json'
},
json={
'q': 'optimization techniques',
'searchMode': 'hybrid',
'limit': 10,
'threshold': 0.7,
'filters': {
'and': [
{
'key': 'category',
'operator': 'equals',
'value': 'deep_learning'
},
{
'key': 'difficulty',
'operator': 'less_than_or_equal',
'value': 3
}
]
}
}
)
data = response.json()
```
</CodeGroup>
## Hybrid search with reranking
Use reranking to improve result relevance in hybrid search:
<CodeGroup>
```bash cURL
curl -X POST https://api.supermemory.ai/v4/search \
-H "Authorization: Bearer YOUR_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"q": "best practices for model training",
"searchMode": "hybrid",
"limit": 10,
"threshold": 0.7,
"rerank": true,
"include": {
"documents": true,
"chunks": true
}
}'
```
```javascript JavaScript
const response = await fetch('https://api.supermemory.ai/v4/search', {
method: 'POST',
headers: {
'Authorization': 'Bearer YOUR_API_KEY',
'Content-Type': 'application/json'
},
body: JSON.stringify({
q: 'best practices for model training',
searchMode: 'hybrid',
limit: 10,
threshold: 0.7,
rerank: true,
include: {
documents: true,
chunks: true
}
})
});
const data = await response.json();
```
```python Python
import requests
response = requests.post(
'https://api.supermemory.ai/v4/search',
headers={
'Authorization': 'Bearer YOUR_API_KEY',
'Content-Type': 'application/json'
},
json={
'q': 'best practices for model training',
'searchMode': 'hybrid',
'limit': 10,
'threshold': 0.7,
'rerank': True,
'include': {
'documents': True,
'chunks': True
}
}
)
data = response.json()
```
</CodeGroup>
## Processing hybrid search results
Here's how to handle both memory and chunk results:
```javascript
const response = await fetch('https://api.supermemory.ai/v4/search', {
method: 'POST',
headers: {
'Authorization': 'Bearer YOUR_API_KEY',
'Content-Type': 'application/json'
},
body: JSON.stringify({
q: 'machine learning concepts',
searchMode: 'hybrid',
limit: 10
})
});
const data = await response.json();
// Process results based on type
data.results.forEach(result => {
if (result.memory) {
// This is a memory result
console.log('Memory:', {
id: result.id,
content: result.memory,
similarity: result.similarity,
metadata: result.metadata
});
} else if (result.chunk) {
// This is a chunk result
console.log('Chunk:', {
id: result.id,
content: result.chunk,
similarity: result.similarity,
document: result.documents[0]?.title || 'Unknown'
});
}
});
```
## When to use hybrid search
Use hybrid search when:
- **Incomplete memories**: Your memories might not cover all relevant information
- **Document-heavy content**: You have large documents that haven't been fully processed into memories
- **Exploratory search**: You want to discover information across all available sources
- **Fallback mechanism**: You want to ensure results even when memories are sparse
Use memories-only search when:
- **Structured data**: You only want curated, structured memory entries
- **Performance**: You need the fastest possible search (hybrid adds ~50-100ms)
- **Memory-first approach**: Your application relies primarily on memories
## Best practices
1. **Set appropriate thresholds**: Use higher thresholds (0.75-0.85) for hybrid search to ensure quality
2. **Include document metadata**: Always include documents to understand chunk sources
3. **Handle both result types**: Check for `memory` or `chunk` fields in your code
4. **Use reranking for quality**: Enable reranking when result quality is critical
5. **Filter appropriately**: Use filters to narrow down results from both sources
6. **Monitor performance**: Hybrid search is slightly slower due to parallel chunk search
## Related
<CardGroup cols={2}>
<Card title="Memory search" icon="brain" href="/search/examples/memory-search">
Examples of memories-only search
</Card>
<Card title="Parameters" icon="sliders" href="/search/parameters">
All search parameters
</Card>
<Card title="Response schema" icon="brackets-curly" href="/search/response-schema">
Understanding responses
</Card>
<Card title="Filtering" icon="filter" href="/search/filtering">
Advanced filtering
</Card>
</CardGroup>