Overview The Search API performs semantic search across all your indexed content in Khoj. It uses embeddings to find relevant information based on meaning, not just keyword matching.
Search Content Search for content across your knowledge base.
curl "https://app.khoj.dev/api/search?q=machine+learning&n=10&t=all" \ -H "Authorization: Bearer YOUR_API_TOKEN"
import requests headers = { "Authorization" : "Bearer YOUR_API_TOKEN" } params = { "q" : "machine learning" , "n" : 10 , "t" : "all" } response = requests.get( "https://app.khoj.dev/api/search" , params = params, headers = headers ) results = response.json()
const params = new URLSearchParams ({ q: 'machine learning' , n: 10 , t: 'all' }); const response = await fetch ( `https://app.khoj.dev/api/search? ${ params } ` , { headers: { 'Authorization' : 'Bearer YOUR_API_TOKEN' } } ); const results = await response . json ();
Request Number of results to return (1-100)
Content type to search. Options:
all - Search all content typesorg - Org-mode filesmarkdown - Markdown filespdf - PDF documentsplaintext - Plain text filesimage - Imagesdocx - Word documentsgithub - GitHub repositoriesnotion - Notion pages Whether to rerank results using cross-encoder for improved relevance
Maximum semantic distance for results (0.0-1.0). Lower values return more similar results.
Whether to deduplicate similar results
Client identifier for telemetry (e.g., “web”, “obsidian”, “emacs”)
Response Returns an array of search results ordered by relevance.
The content snippet or entry text
Relevance score (lower is more relevant)
Additional metadata about the result Compiled/formatted content
Section heading if applicable
Example Response [ { "entry" : "Machine learning is a subset of artificial intelligence..." , "score" : 0.234 , "additional" : { "file" : "notes/ai/machine-learning.md" , "compiled" : "# Machine Learning \n\n Machine learning is a subset..." , "heading" : "Introduction" } }, { "entry" : "Deep learning uses neural networks with multiple layers..." , "score" : 0.312 , "additional" : { "file" : "notes/ai/deep-learning.md" , "compiled" : "## Deep Learning \n\n Deep learning uses neural networks..." , "heading" : "Overview" } } ]
Search Types Khoj supports searching across different content types:
All Content Search across all indexed content:
curl "https://app.khoj.dev/api/search?q=productivity&t=all" \ -H "Authorization: Bearer YOUR_API_TOKEN"
Markdown Files Search only markdown documents:
curl "https://app.khoj.dev/api/search?q=project+notes&t=markdown" \ -H "Authorization: Bearer YOUR_API_TOKEN"
PDF Documents Search only PDF files:
curl "https://app.khoj.dev/api/search?q=research+papers&t=pdf" \ -H "Authorization: Bearer YOUR_API_TOKEN"
GitHub Repositories Search across indexed GitHub repos:
curl "https://app.khoj.dev/api/search?q=api+endpoint&t=github" \ -H "Authorization: Bearer YOUR_API_TOKEN"
Notion Pages Search Notion content:
curl "https://app.khoj.dev/api/search?q=meeting+notes&t=notion" \ -H "Authorization: Bearer YOUR_API_TOKEN"
Advanced Search Options Reranking for Better Results Enable reranking to improve result quality using a cross-encoder model:
curl "https://app.khoj.dev/api/search?q=query&r=true" \ -H "Authorization: Bearer YOUR_API_TOKEN"
Reranking uses a more sophisticated model to reorder results, trading some speed for better relevance.
Filtering by Distance Limit results to those within a semantic distance threshold:
curl "https://app.khoj.dev/api/search?q=query&max_distance=0.3" \ -H "Authorization: Bearer YOUR_API_TOKEN"
Lower distance values mean more similar content. Typical ranges:
0.0 - 0.3: Very similar content0.3 - 0.5: Moderately similar0.5 - 0.7: Somewhat related> 0.7: Less related
Disable Deduplication Get all matching results including similar duplicates:
curl "https://app.khoj.dev/api/search?q=query&dedupe=false" \ -H "Authorization: Bearer YOUR_API_TOKEN"
Use Cases Personal Knowledge Base Search your personal notes and documents:
query = "How do I configure my development environment?" results = search( q = query, n = 5 , t = "markdown" )
Research Assistant Find relevant passages in research papers:
query = "neural network architectures for NLP" results = search( q = query, n = 10 , t = "pdf" , r = True )
Code Search Search through your GitHub repositories:
query = "authentication middleware implementation" results = search( q = query, n = 15 , t = "github" )
Meeting Notes Find information from past meetings:
query = "Q4 roadmap discussion" results = search( q = query, n = 10 , t = "notion" )
Error Handling 401 Unauthorized { "detail" : "Not authenticated" }
Ensure you’re including a valid API token in the Authorization header.
400 Bad Request { "detail" : "Invalid search type specified" }
Check that your search type (t parameter) is one of the supported values.
500 Internal Server Error { "detail" : "Search failed due to internal error" }
The search index may be unavailable. Try again later or contact support.
Best Practices
Use Specific Queries More specific queries return more relevant results. Instead of “python”, try “python async await patterns”.
Choose Right Content Type Filter by content type when you know where information should be (e.g., code in GitHub, docs in Markdown).
Enable Reranking Selectively Use reranking for important queries where accuracy matters more than speed.
Adjust Result Count Start with 5-10 results. Increase if needed, but more results means more data to process.
Rate Limits
Free tier : 50 searches per dayPremium tier : Unlimited searches
Rate limits reset daily at midnight UTC.
Next Steps
Update Content Learn how to index new content for search
