Ida 26.01 is now generally available. This release is built around a single theme: visibility. Visibility into how Ida reasons, where your agent has gaps, and exactly what happens when a user asks a question.
Highlights
AI Thinking Transparency
As Ida has become an agentic assistant, a natural question has followed: what is it actually doing? In 26.01, Ida answers that in real time. When operating in agentic mode, Ida now surfaces its planning process, explaining which tools it’s selecting, why it’s selecting them, and how it’s reasoning through to a response before it delivers one. For end users, this builds trust in answers that come from a multi-step process. For admins, it makes Ida’s behavior auditable and predictable. Those two things matter a lot when you’re deploying AI at scale.
Ida Sources Quality Report
Good answers require good data and content, and now you can see exactly where the gaps are. The new Sources Quality Report surfaces themes from low-confidence responses, giving admins a structured view of which knowledge areas need attention. Rather than reacting to user complaints after the fact, teams can proactively prioritize content improvements based on what Ida actually struggles with. It’s a direct line from AI performance data to content strategy.
Ida Convo Tracer
The Convo Tracer is a new diagnostic tool that gives admins full visibility into any interaction. Enter an utterance and see which knowledge chunks were retrieved, which tools were invoked, and how Ida reasoned through to its answer, step by step. Whether you’re diagnosing unexpected behavior, validating a content update, or just trying to understand why Ida responded a certain way, the Convo Tracer cuts through the guesswork.
Full Release
Enhancements
- AI verbose thinking: Ida now explains its planning, tool selection, and reasoning in real time during agentic interactions
- Updated web channel: Refreshed web channel with tool checkpointing and verbose thinking output support
- Sources Quality Report: New admin report surfacing themes from low-confidence responses to guide content improvements
- Ida Convo Tracer: New diagnostic tool to inspect knowledge retrieval, tool use, and step-by-step reasoning for any utterance
- Support for Claude as LLM provider: Anthropic’s Claude is now available as a language model provider alongside existing options
- Decoupled LLM and Embedding providers: LLM and embedding providers can now be configured independently, giving teams more flexibility in how they deploy and tune AI capabilities
- Tool popularity report: New report tracking which tools are invoked most frequently across interactions
- Programmatic chat access control: A new flag allows integrations to gate chat access until a prerequisite action is completed
- Error code report: New aggregated error report showing error types and volumes over a configurable date range
- Knowledge Web Agent: Improved automatic CSS selector generation for web content ingestion reducing noise and pinpointing the most valuable parts of any web page
Bug Fixes
- Various fixes to conversation log layout, alignment, and timestamp display
- Resolved an encoding issue affecting chat input display in the conversation log viewer
- Various styling and layout fixes across the admin console
We’re confident 26.01 gives your team the tools to operate Ida with greater clarity and confidence. As always, thank you for the continued feedback that shapes each release. We look forward to seeing what you do with it.
Until next time,
The Ida Team
