Apr 25, 2025
AI Axis v1.0.1 introduces in-chat attachment support, smarter tool planning, and advanced image generation for more interactive and accurate conversations. Agents now remember past chats for better continuity and stream step-by-step updates on complex tasks. Enhanced onboarding and profiling enable deeper personalization, while improved error handling and output quality ensure clearer, more reliable responses. The focus shifts from speed to more intelligent and context-aware interactions.
New Features and Enhancements for the AI Axis Release
In-Chat Attachment Support
Agents now support file uploads (images and text).
Attachments are parsed using LLM-driven descriptions and semantic extraction.
Context from uploaded files can be referenced live in conversations for deeper, more accurate interactions.
Enhanced Long-Term Memory
Agents now recall relevant past conversations using vector-based similarity, not just recent turns.
This enables more coherent answers across sessions, better follow-up support, and continuity in long-term discussions.
Live Agent Streaming
Complex requests now stream step-by-step updates as they progress.
Streaming improves transparency and allows users to see how the system is reasoning through a request.
Smarter Tool Planning
Agents now evaluate which tools to use based on the prompt context.
A dedicated agent has been added to generate a linear tool chain that makes sense instead of deeply nesting it in the supervisor runner.
This reduces unnecessary chains and enables more accurate tool invocation for multi-step reasoning e.g. Get details, then generate an image based on details.
Advanced Image Generation
Agents can now create detailed visuals using both:
OpenAI GPT-Image-1 for complex, layout-aware imagery
Stable Diffusion for fast and creative outputs
Web Search with Contextual Awareness
Enabled agents now perform structured web searches, extracting and summarizing real content instead of returning links.
Especially effective for client lookups or competitive insights.
User Experience Improvements
Guided Onboarding Flow
New conversational onboarding collects:
Name
Role
Company website
This generates user-specific corpora, enabling better personalization and immediate agent context.
Customer Profiling & ICP Generation
Agents now generate Ideal Customer Profiles using:
User metadata
Search results
Document enrichment
Chat Interface
Sessions now include:
Auto-generated titles based on topics
Assistant identifiers for multi-agent systems
More robust feedback storage and tracking. No feedback into the void anymore.
Output Quality Improvements
Final Response Overhaul
Final responses are now optimized for clarity, flow, and conversational relevance.
Streaming delivery allows long outputs to arrive in readable chunks, improving comprehension.
Robust Error Handling
Agents now recover from backend issues gracefully.
Agents don't over focus on failed tool calls but work with what they have.
Users receive interpretable explanations and continued dialog flow, not abrupt failures.
Intelligent Behavior, Not Just Speed
This release does not prioritize latency reduction. Instead, it focuses on more intelligent, nuanced reasoning to improve capability and clarity. It also provides performance improvements on the following:
Multi-part inputs
Long-running tasks
Contextually layered conversations (Longer threads) - Conversation over 1 task per chat