Search
clearception.ai's search finds relevant passages across all your uploaded documents, synced workspaces, and connected databases using plain English — no keywords or exact phrasing required.
Opening Search
Go to app.clearception.ai/search or click Search in the sidebar.
How to Search
- Type a natural language query in the search bar (e.g., "meeting notes about Q3 budget" or "customers with overdue invoices").
- Results stream in ranked by semantic relevance.
- Click any result to expand the full passage and see which document or database it came from.
- Use the source filter to narrow results to a specific source (local files, Notion, GitHub, or a connected database).
Knowledge Sources
Search queries across all indexed sources simultaneously:
| Source | How to add |
|---|---|
| Uploaded files | Click Add Documents on the Search page or upload via Data Sources |
| Notion | Connect Notion via Settings → Integrations |
| GitHub | Index a repository via Settings → Integrations |
| Connected databases | Add database connections via Settings → Database Connections — handled by the DeepQuery Agent |
Supported File Formats
| Format | Extension(s) |
|---|---|
.pdf | |
| Markdown | .md |
| Plain text | .txt |
| Org-mode | .org |
| Word documents | .docx |
Query Filters
Append special filters to your query to narrow results:
| Filter | Example | Effect |
|---|---|---|
file: |
budget file:q3-report.pdf |
Limit results to a specific file |
date: |
decisions date:2024 |
Limit results to a date range |
word: |
word:profitability |
Require an exact keyword match alongside semantic search |
How It Works
clearception.ai uses a multi-stage retrieval pipeline so synonyms, paraphrasing, and context all work — you don't need to remember exact wording from your notes.
- Indexing — when you add a document or database, a bi-encoder embedding model creates meaning vectors for each chunk of text.
- Retrieval — your search query is converted into a meaning vector, and the closest chunks are retrieved by comparing vectors (semantic similarity).
- MMR deduplication — near-duplicate results are filtered out to maximize coverage across different topics and documents.
- Re-ranking — a cross-encoder model re-ranks the top results for final accuracy before display.
Database Search via DeepQuery
When you have a database connection set up, Search automatically routes relevant queries to the DeepQuery Agent. DeepQuery combines vector similarity search with AI-generated SQL to return the most relevant records and schema context. You can also interact with DeepQuery directly from Chat for more complex multi-step database queries.
Search also powers the context used in Chat — every time you ask a question, clearception.ai runs a search behind the scenes to find your most relevant notes before generating a response.
Downloading Results
Click the Download button on the search results page to export the current result set. Results are exported as a structured file for further analysis or sharing.