microsoft / microsoft/simplechat

Enable core chat orchestration to use Workspace Search, Analyze, and Compare

Open
#994 0 comments 0 reactions 1 assignee Claimed by @paullizer View on GitHub
enhancement
Dominant language
Python
Stars
152
Forks
116
Avg merge
7h 7m
Merged PRs (30d)
122

Description

## Summary

Build Workspace document action orchestration into the core Semantic Kernel chat agent/action layer used by both streaming and non-streaming Chat.

Today, the Simple Chat action can search workspace knowledge, but it does not support Analyze or Compare, and it does not have intent-selection logic to decide whether a user request should use Search, Analyze, Compare, or normal Q&A.

This issue proposes making that orchestration a core Chat capability so:

- Normal Chat users can get better document reasoning without manually selecting an agent.
- Agents using assigned knowledge plus the Simple Chat action automatically gain the same Workspace-aware behavior.
- Workflows built on top of agents can also benefit from Search, Analyze, Compare, and clarifying-question behavior.

## User Value

Users currently need to understand the difference between Workspace mode, Analyze, Search, Compare, and normal Chat. That creates a multi-step flow and requires mode switching.

The improved experience should allow users to ask naturally, while the core chat agent determines whether it needs to search, analyze, compare, ask a clarifying question, or produce a final answer.

This also makes agents more powerful without requiring every agent to reimplement document-action logic separately.

## Proposed Behavior

- The core Semantic Kernel chat agent/action layer can use Workspace document actions.
- The same capability works in both streaming and non-streaming Chat.
- The Simple Chat action can invoke Search, Analyze, and Compare where appropriate.
- The chat agent can decide which action to use based on user intent:
- Search for targeted retrieval or fact finding.
- Analyze for deeper extraction, summarization, structured analysis, or report-building.
- Compare when the user asks about differences, similarities, gaps, conflicts, or tradeoffs between documents.
- General Q&A when no Workspace action is needed.
- The chat agent can ask clarifying questions before choosing an action when user intent is ambiguous.
- Clarifying questions can include structured chat UI inputs, such as buttons or selectable document choices.
- For Compare, the chat flow can ask which documents should be compared when the request does not clearly identify them.
- Agents with assigned knowledge and the Simple Chat action inherit this capability.
- Workflows using those agents can also use this capability.

## Acceptance Criteria

- [ ] Core Semantic Kernel chat orchestration supports Search, Analyze, and Compare over workspace documents.
- [ ] The behavior works in streaming Chat.
- [ ] The behavior works in non-streaming Chat.
- [ ] The Simple Chat action can invoke Analyze and Compare, not only Search.
- [ ] Intent-selection logic chooses Search, Analyze, Compare, or general Q&A based on the user request.
- [ ] The chat agent can ask clarifying questions when intent or document selection is ambiguous.
- [ ] Chat UI supports structured clarification inputs such as selectable buttons or document choices.
- [ ] Compare flow can request or infer which documents should be compared.
- [ ] Users can benefit from this behavior without selecting a custom agent.
- [ ] Agents using assigned knowledge plus the Simple Chat action inherit the same behavior.
- [ ] Workflows built on agents can use the same capability.
- [ ] Workspace document permissions and selected-document scoping are preserved.
- [ ] Functional tests cover streaming, non-streaming, agent, and workflow-relevant paths.

## Notes

This should not be implemented as a one-off Workspace Document Analysis Agent feature only.

The preferred direction is to make Workspace document action selection a core capability of the chat orchestration layer. A specialized agent can still be created later, but the base Chat experience and Simple Chat action should both gain the ability first.

Current gap:

- Simple Chat action supports Search.
- Simple Chat action does not support Analyze.
- Simple Chat action does not support Compare.
- Chat orchestration does not currently choose between Search, Analyze, Compare, and Q&A based on user intent.
- Chat UI may need new clarification controls for intent selection and document selection, especially for Compare.

Contributor guide

Open the contributing guide

Assessment

This issue has not been assessed yet.

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.