Sentinel Coder One Studio now includes the Foundry IQ comparison proof assets directly in the GitHub/Marketplace package so the evidence links resolve for users. The concise comparison remains: no-IQ answer 67/100 vs Foundry-IQ grounded answer 100/100, with 9 live sources retrieved from the real Knowledge Base.
| Proof item | Result |
|---|---|
| Microsoft IQ layer | Foundry IQ |
| Runtime path | Azure AI Search Knowledge Base retrieve |
| Knowledge Base | sentinel-coder-iq-kb on qubitpage-srch |
| Sources retrieved | 9 |
| No-IQ answer score | 67 / 100 |
| Foundry-IQ answer score | 100 / 100 |
| Evidence | docs/foundry-iq/FOUNDRY_IQ_VS_NO_FOUNDRY_INFOGRAPHIC.html |
Benefit: without Foundry IQ, the agent gives a generic implementation plan. With Foundry IQ, Sentinel Coder grounds the answer in current Sentinel Coder One release/docs evidence, reports retrieved sources, and avoids inventing product-specific facts.
The Microsoft IQ runtime no longer spawns Azure CLI or shell commands from the VS Code extension. Foundry IQ Knowledge Base retrieve endpoints now require a bearer/JWT token supplied through sentinelCoder.microsoftIq.apiKeyEnv (default MICROSOFT_IQ_BEARER_TOKEN). Azure CLI is used only by local provisioning/test scripts, not by Marketplace runtime code.
Sentinel Coder One Studio now defaults to the verified Microsoft Foundry IQ Knowledge Base retrieve endpoint on qubitpage-srch: sentinel-coder-iq-kb. The runtime supports the documented 2026-04-01 retrieve schema with intents and knowledgeSourceParams, and can obtain Azure Search bearer tokens via Azure CLI for local desktop testing without storing tokens in source.
Sentinel Coder One Studio now integrates the required Microsoft IQ layer for the Microsoft Agents League Hackathon through Foundry IQ backed by Azure AI Search.
- Real Foundry IQ backend:
sentinel-coder-iqindex on Azure AI Search servicestanin resource groupfoundryIQsource. - Real grounding in prompts: when enabled, Sentinel queries Foundry IQ before model calls and injects retrieved enterprise/project context into the agent prompt.
- Visible setup path: Foundry IQ appears in the real inline sidebar UI and Settings overlay with Enable, layer, endpoint, API env var, timeout, max-query, Save, and Test controls.
- Verified endpoint: extension-module smoke test returns Foundry IQ context from the Azure Search index using
MICROSOFT_IQ_BEARER_TOKENwithout printing secrets. - Chat visibility fixes: latest assistant output, restored sessions, and Foundry IQ test results are pinned to visible chat output while preserving manual scroll behavior.
This submission explicitly uses Microsoft IQ / Foundry IQ:
| Requirement | Sentinel Coder One implementation |
|---|---|
| IQ layer | Foundry IQ |
| Backend | Azure AI Search index sentinel-coder-iq |
| Endpoint setting | sentinelCoder.microsoftIq.endpoint |
| Secret handling | Bearer/JWT token read from MICROSOFT_IQ_BEARER_TOKEN; never hardcoded |
| Agent behavior | Retrieves project/enterprise knowledge and injects it into coding-agent prompts |
| UI behavior | Foundry IQ setup and test controls are visible in Sentinel Coder One Studio |
Multi-provider chat | Categorized live model selector | Single-model full-capability mode | Opt-in Agentic orchestration | Microsoft Foundry IQ grounding | Azure/OpenAI/Anthropic/Groq/OpenRouter/Ollama and OpenAI-compatible providers | Sora video | Image/audio/document Studio | VS Code Web + Remote Tool Bridge
Built by QubitPage Research | MIT licensed
GitHub repository | Contributing guide | Issues and feature requests
- Fixed the real rendered Sentinel Settings overlay: Microsoft IQ / Foundry IQ opens a real inline
settings-iqpane instead of a blank Settings page. - Added visible Enable, layer, endpoint, API env var, timeout, max-query, Save, and Test controls to the actual inline webview HTML in
src/sidebarProvider.ts. - Verified Azure AI Search-backed Foundry IQ through
MICROSOFT_IQ_BEARER_TOKEN.
- Fixed the real inline chat webview layout after Foundry IQ visibility work.
- Kept a full-width Foundry IQ banner below the toolbar so it cannot cover or break the chat header/output area.
- Fixed the actual rendered webview path: Foundry IQ appears in inline
src/sidebarProvider.tsHTML, not only in unusedmedia/sidebar.html. - Added an always-visible Microsoft Foundry IQ banner directly below the top toolbar with Setup and Test buttons.
- Configured real Azure AI Search endpoint for Foundry IQ retrieval.
- Verified extension-module smoke test with real endpoint and non-secret API-key environment variable.
- Microsoft IQ / Foundry IQ uses the real Azure AI Search-backed
sentinel-coder-iqindex by default. - Azure AI Search Foundry IQ Knowledge Base retrieve endpoints are called with bearer/JWT authentication from
MICROSOFT_IQ_BEARER_TOKEN.
Important Foundry IQ settings:
| Setting | Purpose |
|---|---|
sentinelCoder.microsoftIq.enabled |
Enables Microsoft IQ grounding. |
sentinelCoder.microsoftIq.layer |
Use foundry-iq for this hackathon submission. |
sentinelCoder.microsoftIq.endpoint |
Azure AI Search Foundry IQ endpoint. |
sentinelCoder.microsoftIq.apiKeyEnv |
Environment variable name containing the bearer/JWT token, normally MICROSOFT_IQ_BEARER_TOKEN. |
sentinelCoder.microsoftIq.timeoutMs |
Request timeout. |
sentinelCoder.microsoftIq.maxQueryChars |
Max query size sent to Foundry IQ. |
Other settings are available in VS Code Settings under Sentinel Coder One Studio.
- Sentinel is bring-your-own-key.
- API keys are read from environment variables or VS Code SecretStorage/user settings where applicable.
- Foundry IQ credentials are not hardcoded into source or Marketplace artifacts.
- Tool results are summarized/collapsed in the UI to avoid flooding chat with raw command output.