Features

A comprehensive look at Ida's capabilities—from real-time transcription to intelligent search and AI-powered recommendations.

Real-Time Transcription

Ida captures audio and converts it to text in real-time, powered by Groq Whisper (whisper-large-v3) for fast, accurate transcription.

Audio Sources

  • System Audio — Capture sound from any application (Zoom, Meet, Teams, etc.)
  • Microphone — Record your own voice directly
  • Both — Capture meeting audio and your microphone simultaneously

Transcription Features

  • Speaker Identification — Automatically labels different speakers
  • Timestamps — Each segment includes timing information
  • Real-time Display — See transcription as it happens
  • Custom Vocabulary — Add domain-specific terms for better accuracy
Custom Vocabulary

Add company names, product terms, technical jargon, and people's names to improve transcription accuracy. Open Settings → Custom Vocabulary to add terms.

Vocabulary Categories

  • Company Terms — Organization names, department names, project codenames
  • Product Names — Your products and services
  • Technical Terms — Acronyms, API names, frameworks (Kubernetes, OAuth, GraphQL)
  • People Names — Colleagues and frequent contacts

Document Indexing

Build your knowledge base by indexing documentation from multiple sources. Ida supports several ingestion methods:

Web Crawler

Automatically crawl and index entire documentation websites:

  • Enter a starting URL (e.g., https://docs.example.com/getting-started)
  • Set a page limit (default: 100, max: 1000)
  • Crawler discovers linked pages within the same domain
  • Content is extracted, chunked, and embedded for search

Crawler Settings

Setting Default Description
Max Pages 100 Maximum number of pages to crawl
Delay 300ms Delay between requests to be respectful
Stay in Path Yes Only follow links under the starting URL path

Local Files

Index Markdown files from your local filesystem:

  • Select a folder containing .md files
  • Recursively indexes all Markdown documents
  • Preserves document structure and headings

Pasted Text

Quickly index snippets of text:

  • Copy text from any source
  • Paste into Ida with a descriptive title
  • Instant indexing for immediate use

Semantic Chunking

Ida uses intelligent chunking to split documents into meaningful pieces that preserve context:

How It Works

  • Sentence-aware splitting — Never breaks mid-sentence
  • Structure preservation — Respects headings and paragraph boundaries
  • Code block protection — Keeps code blocks intact
  • Context overlap — Adjacent chunks share sentences for continuity

Chunk Settings

Parameter Default Description
Target Size ~400 tokens Ideal chunk size for retrieval
Max Size ~600 tokens Hard limit before forced split
Min Size ~100 tokens Prevents tiny, context-poor chunks
Overlap 2 sentences Shared context between chunks

Hybrid Search

Ida combines two search methods for optimal results:

Vector Search (Semantic)

  • Uses OpenAI text-embedding-3-small embeddings
  • Understands meaning and context
  • Finds relevant content even with different terminology
  • Great for natural language queries

Keyword Search (BM25)

  • Uses SQLite FTS5 full-text search
  • Excellent for exact matches and technical terms
  • Handles API names, function names, and acronyms
  • Porter stemming for word form normalization

Score Fusion

Results are combined using Reciprocal Rank Fusion (RRF): 

combined_score = (vector_score × 0.7) + (bm25_score × 0.3)

// If both methods find the result, apply 1.2× boost
if (found_in_both) combined_score *= 1.2
Why Hybrid Search?

Semantic search alone can miss exact technical terms. Keyword search alone misses conceptual matches. Together, they provide the best of both worlds.

Keyword Triggers (Instant FTS)

Keyword triggers enable instant documentation retrieval when specific keywords are mentioned in conversation. Instead of waiting for semantic analysis, Ida runs an immediate full-text search (FTS) when it detects configured keywords.

How Keyword Triggers Work

  1. You configure keywords to watch for (e.g., "google calendar", "gcal", "calendar sync")
  2. You define an FTS query to run when those keywords are detected
  3. During transcription, Ida monitors for these keywords in real-time
  4. When a match is found, it immediately runs the FTS query across your active vaults
  5. Results appear instantly—typically under 100ms—much faster than semantic search

Creating Keyword Triggers

Open Settings → Documentation → Instant Recommendations to manage keyword triggers:

  • Manual Creation — Add keywords and an FTS query manually
  • AI Suggestions — Click "Generate Suggestions" to get AI-powered trigger ideas based on your indexed documentation
  • Enable/Disable — Toggle individual triggers without deleting them
  • Edit — Modify keywords or queries for existing triggers

Example Trigger

Keywords FTS Query Use Case
google calendar, gcal, calendar sync google calendar oauth sync Instantly surfaces calendar integration docs when calendar is mentioned
authentication, auth, login authentication oauth JWT token Shows auth-related documentation when discussing login flows
API rate limit, throttle rate limit throttle quota Provides rate limiting info when performance is discussed

FTS Query Syntax

Keyword trigger queries use SQLite FTS5 syntax for powerful search capabilities:

  • Multiple terms — Space-separated terms: oauth token refresh
  • OR queries — Use OR for alternatives: authentication OR oauth
  • Phrase search — Use quotes for exact phrases: "API key rotation"
  • Prefix search — Use * for prefix matching: auth* matches "authentication", "authorize", etc.
When to Use Keyword Triggers

Keyword triggers are perfect for technical terms, product names, and specific topics that have clear documentation. They work best when:

  • You have well-organized documentation on specific topics
  • Certain keywords reliably indicate what documentation is needed
  • Speed is critical—you want instant results during fast-moving conversations

AI-Generated Suggestions

Need help getting started? Use the AI suggestion feature to analyze your indexed documentation and generate smart trigger recommendations:

  1. Index some documentation in your vault
  2. Go to Settings → Documentation → Instant Recommendations
  3. Click "Generate Suggestions"
  4. Review the 3 suggested triggers based on your content
  5. Click "Add" to activate any suggestions you like

Keyword Triggers vs. Semantic Recommendations

Feature Keyword Triggers (FTS) Semantic Recommendations
Speed ~100ms (instant) ~300ms
Accuracy Exact matches for known terms Conceptual understanding
Setup Requires configuration Works automatically
Best For Technical terms, product names, specific topics Natural language, exploratory questions

Both systems work together—keyword triggers provide instant results when keywords match, otherwise the standard semantic recommendation system takes over.

AI Recommendations

During meetings, Ida automatically surfaces relevant documentation based on the conversation:

How Recommendations Work

  1. Transcription is analyzed for key topics and questions
  2. Hybrid search finds the most relevant chunks
  3. AI (Groq LLaMA 3.3 70B) synthesizes an answer with sources
  4. Recommendations appear in ~300ms

Recommendation Features

  • Source Citations — See exactly where information came from
  • Relevance Scoring — Most relevant docs appear first
  • Pinning — Save important recommendations for later
  • Click to Expand — View full context of any source

Chat Interface

Ask questions about your documentation and get AI-powered answers:

Chat Features

  • RAG-powered — Answers grounded in your indexed documentation
  • Source Citations — Every answer includes relevant sources
  • Context Aware — Chat considers current session's transcription
  • History — Conversation history is saved per session

Tips for Better Results

  • Be specific in your questions
  • Reference topics from the transcription for context
  • Ask follow-up questions to drill deeper
  • Use the pin feature to save important answers

Notes

Take and organize notes at multiple levels:

Session Notes

  • Private notes for each session
  • Rich text editing with Markdown support
  • Auto-saved as you type

Account Notes

  • Persistent notes tied to an account (not a session)
  • Great for ongoing context about a client or company
  • Visible across all sessions with that account

Themes

Ida includes three color themes to match your preference:

Ida (Default)

Deep navy gradient with bright blue accents. Modern and focused.

Dark

High contrast dark theme inspired by VS Code. Maximum readability.

Light

Clean white theme for well-lit environments and presentations.

Change themes in Settings → Appearance. Your preference is saved and persists across sessions.

Next Steps