Skip to main content
Version: v0.2 (Next)

Multimodal AI Chat & Interactions

The ActonOS Chat Interface provides an interactive conversational environment supporting single-agent dialogues, multi-agent collaborations, document and image analysis, live tool execution cards, and real-time streaming deltas via Server-Sent Events (SSE).


1. Starting a Conversation​

  1. Navigate to Chat in the left sidebar.
  2. In the top-left conversation rail, click + New Chat.
  3. Select the target agent from the Agent Selector dropdown (e.g., General Assistant, Research Agent, Coder Swarm).
  4. Type your message into the rich prompt editor and press Enter (or Shift + Enter for multi-line inputs).

2. Multimodal Attachments & Document Ingestion​

ActonOS features a client-side document and image preprocessor that parses attachments before ingestion:

File TypeSupported FormatsProcessing Method
Text & Code.txt, .md, .py, .go, .js, .ts, .json, .yaml, .sqlUTF-8 plain text extraction with syntax highlighting
Documents.pdf, .docx, .docExtracted into clean structured Markdown text with page markers
Tabular Data.csv, .tsv, .xlsx, .xlsConverted into Markdown data tables with schema extraction
Visual Media.png, .jpg, .jpeg, .webp, .svgResized & optimized for vision models (Claude 3.7 / GPT-4o / Gemini 2.0 Flash)

To attach files:

  • Click the πŸ“Ž Paperclip icon in the composer.
  • Or simply drag and drop files directly onto the chat window.

3. Real-Time Tool Execution Cards​

When an agent invokes external tools (e.g., executing a Python script, browsing a website, or querying an SQLite database), ActonOS renders an interactive Tool Execution Card directly in the message timeline:

β”Œβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”
β”‚ ⚑ native_browser_headless β”‚
β”‚ Target: https://github.com/actonos/actonos β”‚
β”œβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€
β”‚ Status: Completed in 1.42s Β· Extracted 2,840 bytes β”‚
β”‚ [β–Ά Click to inspect JSON input & output parameters] β”‚
β””β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”˜
  • Live Status Indicator: Shows whether the tool is Executing, Completed, or Failed.
  • Parameter Inspection: Click the disclosure chevron to review the exact JSON arguments passed to the tool and the returned output payload.
  • Safety Flags: If a tool requires human confirmation, the card turns amber with direct Approve and Reject controls.

4. Message Branching & Context Management​

  • Copy Code Blocks: One-click copy buttons on all code snippets with automatic language detection.
  • Regenerate / Fork: Click the regenerate button on any agent response to switch to a different model in the Cascade router or adjust the prompt.
  • Context Compaction: For long multi-turn sessions, ActonOS automatically creates semantic snapshots in context_snapshots to preserve token budgets while maintaining critical conversational memory.