feat: onboard new sample data analyst agent v2

.config/samples-config-v3.json

@@ -937,6 +937,30 @@
             "thumbnailPath": "images/screen009.jpg",
             "gifPath": "images/screen009.jpg",
             "suggested": true
+        },
+        {
+            "id": "data-analyst-agent-v2",
+            "shortId": "data-analyst-v2",
+            "onboardDate": "2025-12-23",
+            "title": "Data Analyst Agent v2",
+            "shortDescription": "Natural language interface for data exploration and visualization.",
+            "fullDescription": "This sample demonstrates how to build an AI-powered data analyst agent using Microsoft Teams SDK that can be integrated into Microsoft Teams. It helps users explore and visualize data through natural language conversations and Adaptive Cards charts.",
+            "types": [
+                "Custom Engine Agent"
+            ],
+            "tags": [
+                "TS",
+                "Custom Engine Agent",
+                "Data Visualization",
+                "Adaptive Cards",
+                "LLM SQL",
+                "Microsoft Teams SDK"
+            ],
+            "time": "10 mins to run",
+            "configuration": "Manual configurations required",
+            "thumbnailPath": "assets/demo.gif",
+            "gifPath": "assets/demo.gif",
+            "suggested": false
         }
     ]
-}
+}

.github/PULL_REQUEST_TEMPLATE.md

@@ -0,0 +1,61 @@
+# Pull Request
+
+## Description
+
+<!-- Please provide a brief description of the changes in this PR -->
+
+## Type of Change
+
+- [ ] New sample onboarding (internal - source code in this repo)
+- [ ] New sample onboarding (external - source code in another repo)
+- [ ] Sample update/fix
+- [ ] Documentation update
+- [ ] Validation tool update
+- [ ] Other (please describe)
+
+---
+
+## For New Sample Onboarding
+
+### Checklist
+
+- [ ] I have added/updated the sample entry in `.config/samples-config-v3.json`
+- [ ] I have included a README.md with setup instructions
+- [ ] I have included a thumbnail image with correct aspect ratio (40:23, e.g., 1600×920 or 800×460)
+
+### Validation Results (Required)
+
+> **Important**: You must run the validation tool locally and provide a screenshot of the results.
+
+#### For Internal Samples (source code in this repo)
+
+```bash
+cd validation-tool
+npm install
+node validator.cjs -p ../<your-sample-folder>
+```
+
+#### For External Samples (source code in another repo)
+
+```bash
+cd validation-tool
+npm install
+
+# Clone your sample repo (sparse checkout recommended)
+git clone --filter=blob:none --sparse <your-repo-url>
+cd <repo-name>
+git sparse-checkout set <path-to-sample>
+cd ..
+
+# Run validation
+node validate-external.js <sample-id> ./<repo-name>
+```
+
+
+---
+
+## Related Issues
+
+<!-- Link any related issues here, e.g., "Fixes #123" or "Related to #456" -->
+if any question related to validation, may refer to the [Sample Validation Guide](validation-tool/sample_validation.md)
+if still has questions, may open a issue :)

.github/prompts/onboard-sample.prompt.md

@@ -0,0 +1,69 @@
+---
+mode: agent
+description: "Onboard a sample to Microsoft 365 Agents Toolkit Samples repository"
+tools: ["editFiles", "runInTerminal", "readFile"]
+---
+
+You are onboarding a sample to the Microsoft 365 Agents Toolkit Samples repository.
+
+**CRITICAL: You must DO the work, not explain what to do. Execute each step by reading/writing files and running commands directly. Only ask the user when you truly cannot infer required information.**
+
+The user will provide a sample folder name (e.g., `data-analyst-agent-v2`).
+
+---
+
+## Step 1: Locate and Analyze
+
+1. Find `microsoft-365-agents-toolkit-samples` in workspace
+2. Read sample's `package.json` (TS/JS), `.sln` (C#), or determine it's Python
+3. Check if sample is internal (inside repo) or external
+
+## Step 2: Fix m365agents.yml
+
+1. Read the file (root for TS/JS, or `M365Agent/` for C#)
+2. **DELETE** any `projectId` field if present
+3. **ENSURE** this exists (add if missing):
+   ```yaml
+   additionalMetadata:
+     sampleTag: TeamsFx-Samples:{folder-name}
+   ```
+4. **SAVE the file** with your changes
+
+## Step 3: Fix manifest.json
+
+1. Read `appPackage/manifest.json` (or `M365Agent/appPackage/manifest.json` for C#)
+2. **ENSURE** id field is: `"id": "${{TEAMS_APP_ID}}"`
+3. **SAVE the file** if changed
+
+## Step 4: Update samples-config-v3.json
+
+1. Read `.config/samples-config-v3.json`
+2. Search for entry with matching `id`
+3. If NOT found, **CREATE and INSERT** a new entry:
+   - Infer `title`, `shortDescription`, `fullDescription` from README.md
+   - Infer `types` and `tags` from project content
+   - Set `onboardDate` to today (YYYY-MM-DD)
+   - Set `thumbnailPath` and `gifPath` based on assets folder
+   - For external samples: get `downloadUrlInfo` via `git remote get-url origin`
+4. **SAVE the file** with valid JSON
+
+## Step 5: Run Validation (Internal TS/JS or C# only)
+
+1. Run in terminal:
+   ```bash
+   cd validation-tool && npm install && node validator.cjs -p ../{sample-folder}
+   ```
+2. If validation fails:
+   - **FIX the issues yourself** (resize images, update configs, etc.)
+   - Re-run validation until it passes
+3. Only ask user if you cannot fix an issue automatically
+
+## Step 6: Report Results
+
+Show a brief checklist:
+- [x] or [ ] m365agents.yml has sampleTag, no projectId
+- [x] or [ ] manifest.json uses ${{TEAMS_APP_ID}}
+- [x] or [ ] Entry in samples-config-v3.json
+- [x] or [ ] Validation passed
+
+List any issues that need user action.

data-analyst-agent-v2/README.md

@@ -39,7 +39,7 @@ The Data Analyst Agent v2 transforms how teams interact with data by providing a
 2. Configure environment variables:
     - Open the project in Visual Studio Code
     - The Microsoft 365 Agents Toolkit will automatically generate the required environment files
-    - Update the `.localConfigs` file (or `.localConfigs.playground` if debug with playground) with your Azure OpenAI configuration:
+    - Update the `.env.local.user` file (or `.env.playground.user` if debug with playground) with your Azure OpenAI configuration:
       - `SECRET_AZURE_OPENAI_API_KEY`: Your Azure OpenAI API key
       - `AZURE_OPENAI_ENDPOINT`: Your Azure OpenAI endpoint
       - `AZURE_OPENAI_DEPLOYMENT_NAME`: Your Azure OpenAI deployment name (e.g., gpt-4o)
@@ -152,6 +152,23 @@ Both evaluations will provide detailed feedback including:
 | Date         | Author     | Comments                               |
 | ------------ | ---------- | -------------------------------------- |
 | Oct 31, 2025  | qinzhouxu   | Onboard sample with Microsoft Teams SDK |
+| Dec 16, 2025  | quke      | fix issue and onboard sample           |
+
+## Known Issues
+
+### Service Principal Creation Failure During Local Debug (Admin Accounts Only)
+
+**Impact**: Local debug only. This does not affect debugging in Microsoft 365 Agents Playground or remote deployment to Azure.
+
+**Symptom**: When running local debug with an M365 admin account, you may encounter the following error:
+
+```
+The client application <client-id> is missing service principal in the tenant <tenant-id>.
+```
+
+**Root Cause**: Teams Developer Portal has a known issue where it cannot automatically create a service principal for M365 admin accounts.
+
+**Solution**: Use a non-admin M365 account for local debugging. You can create a regular user account in your M365 tenant and use it for the local debug workflow.
 
 ## Feedback
 We really appreciate your feedback! If you encounter any issue or error, please report issues to us following the [Supporting Guide](https://github.com/OfficeDev/TeamsFx-Samples/blob/dev/SUPPORT.md). Meanwhile you can make [recording](https://aka.ms/teamsfx-record) of your journey with our product, they really make the product better. Thank you!

data-analyst-agent-v2/evals/judge/ac.ts

@@ -48,9 +48,9 @@ consider the submission correct even if titles, labels, colors, or other propert
     const prompt = new ChatPrompt({
         instructions: systemMessage,
         model: new OpenAIChatModel({
-            model: process.env.AOAI_MODEL!,
-            apiKey: process.env.AOAI_API_KEY!,
-            endpoint: process.env.AOAI_ENDPOINT!,
+            model: process.env.AZURE_OPENAI_DEPLOYMENT_NAME!,
+            apiKey: process.env.AZURE_OPENAI_API_KEY!,
+            endpoint: process.env.AZURE_OPENAI_ENDPOINT!,
             apiVersion: '2025-04-01-preview',
         }),
     }).function(
@@ -90,7 +90,31 @@ consider the submission correct even if titles, labels, colors, or other propert
 ************
 [END DATA]`;
             const res = await prompt.send(userPrompt, { autoFunctionCalling: false });
-            const functionCallArgs = res.function_calls?.[0].arguments;
+
+            // Parse function call from content (model returns JSON in content when autoFunctionCalling is false)
+            let functionCallArgs: { result?: boolean; reasoning?: string } | undefined;
+
+            if (res.function_calls?.[0]?.arguments) {
+                functionCallArgs = res.function_calls[0].arguments;
+            } else if (res.content) {
+                // Try to parse JSON-like content without using a complex, backtracking-prone regex
+                try {
+                    const resultMatch = res.content.match(/"result"\s*:\s*(true|false)/);
+                    const reasoningMatch = res.content.match(/"reasoning"\s*:\s*"((?:[^"\\]|\\.)*)"/);
+
+                    if (resultMatch && reasoningMatch) {
+                        const rawReasoning = reasoningMatch[1];
+                        functionCallArgs = {
+                            result: resultMatch[1] === 'true',
+                            reasoning: rawReasoning
+                                .replace(/\\"/g, '"')
+                                .replace(/\\n/g, '\n')
+                        };
+                    }
+                } catch (e) {
+                    // Parsing failed, will use fallback
+                }
+            }
 
             return {
                 result: functionCallArgs?.result || false,

data-analyst-agent-v2/evals/judge/sql.ts

@@ -32,9 +32,9 @@ The submitted answer may either be correct or incorrect. Determine which case ap
     const prompt = new ChatPrompt({
         instructions: systemMessage,
         model: new OpenAIChatModel({
-            model: process.env.AOAI_MODEL!,
-            apiKey: process.env.AOAI_API_KEY!,
-            endpoint: process.env.AOAI_ENDPOINT!,
+            model: process.env.AZURE_OPENAI_DEPLOYMENT_NAME!,
+            apiKey: process.env.AZURE_OPENAI_API_KEY!,
+            endpoint: process.env.AZURE_OPENAI_ENDPOINT!,
             apiVersion: '2025-04-01-preview',
         }),
     }).function(
@@ -73,7 +73,45 @@ The submitted answer may either be correct or incorrect. Determine which case ap
 ************
 [END DATA]`;
             const res = await prompt.send(userPrompt, { autoFunctionCalling: false });
-            const functionCallArgs = res.function_calls?.[0]?.arguments;
+
+            // Parse function call from content (model returns JSON in content when autoFunctionCalling is false)
+            let functionCallArgs: { result?: boolean; reasoning?: string } | undefined;
+
+            if (res.function_calls?.[0]?.arguments) {
+                functionCallArgs = res.function_calls[0].arguments;
+            } else if (res.content) {
+                // Try to parse JSON from content
+                try {
+                    const content = res.content;
+
+                    // First, try to extract a JSON object and parse it directly
+                    const objectMatch = content.match(/\{[\s\S]*\}/);
+                    if (objectMatch) {
+                        const parsed = JSON.parse(objectMatch[0]);
+                        if (typeof parsed.result === 'boolean' && typeof parsed.reasoning === 'string') {
+                            functionCallArgs = {
+                                result: parsed.result,
+                                reasoning: parsed.reasoning
+                            };
+                        }
+                    }
+
+                    // Fallback: use a simpler regex extraction if JSON.parse did not yield result
+                    if (!functionCallArgs) {
+                        const fallbackMatch = content.match(
+                            /"result"\s*:\s*(true|false)[\s\S]*?"reasoning"\s*:\s*"([^"\\]*(?:\\.[^"\\]*)*)"/
+                        );
+                        if (fallbackMatch) {
+                            functionCallArgs = {
+                                result: fallbackMatch[1] === 'true',
+                                reasoning: fallbackMatch[2].replace(/\\"/g, '"').replace(/\\n/g, '\n')
+                            };
+                        }
+                    }
+                } catch (e) {
+                    // Parsing failed, will use fallback
+                }
+            }
 
             return {
                 result: functionCallArgs?.result || false,

data-analyst-agent-v2/evals/skills/adaptive-card-generation.ts

@@ -1,5 +1,6 @@
 import { ACJudge } from '../judge/ac';
 import { AgentEvaluator } from './base-evaluator';
+import { generateChartCard } from '../../src/cards';
 
 new AgentEvaluator({
     evalName: 'ac-eval',
@@ -8,7 +9,65 @@ new AgentEvaluator({
     judge: ACJudge,
     generatePrompt: (tc) =>
         `Create an appropriate visualization for this data: ${JSON.stringify(tc.input_data)}. Please return a single card.\nUse the following type of visualization: ${tc.visualization_type}.`,
-    extractGenerated: (agent) => JSON.stringify(agent.attachments?.[0] ?? {}),
+    extractGenerated: (agent, response) => {
+        // First check if there are actual attachments from function execution
+        if (agent.attachments?.length > 0) {
+            const attachment = agent.attachments[0];
+            return JSON.stringify(attachment.content);
+        }
+
+        // When model returns JSON in content instead of calling functions
+        const content = response.content || '';
+
+        try {
+            // Try format 1: {"name":"functions.generate_card","arguments":{...}}
+            let startIdx = content.indexOf('{"name":"functions.generate_card"');
+            if (startIdx !== -1) {
+                let braceCount = 0;
+                let endIdx = startIdx;
+                for (let i = startIdx; i < content.length; i++) {
+                    if (content[i] === '{') braceCount++;
+                    else if (content[i] === '}') braceCount--;
+                    if (braceCount === 0) {
+                        endIdx = i + 1;
+                        break;
+                    }
+                }
+                const jsonStr = content.substring(startIdx, endIdx);
+                const functionCall = JSON.parse(jsonStr);
+                const { chartType, rows, options } = functionCall.arguments;
+                if (chartType && rows) {
+                    const attachment = generateChartCard(chartType, rows, options);
+                    return JSON.stringify(attachment.content);
+                }
+            }
+
+            // Try format 2: Direct {"chartType":"...","rows":[...],...} at start of content
+            startIdx = content.indexOf('{"chartType"');
+            if (startIdx !== -1) {
+                let braceCount = 0;
+                let endIdx = startIdx;
+                for (let i = startIdx; i < content.length; i++) {
+                    if (content[i] === '{') braceCount++;
+                    else if (content[i] === '}') braceCount--;
+                    if (braceCount === 0) {
+                        endIdx = i + 1;
+                        break;
+                    }
+                }
+                const jsonStr = content.substring(startIdx, endIdx);
+                const { chartType, rows, options } = JSON.parse(jsonStr);
+                if (chartType && rows) {
+                    const attachment = generateChartCard(chartType, rows, options);
+                    return JSON.stringify(attachment.content);
+                }
+            }
+        } catch (e) {
+            // Silent fail
+        }
+
+        return '{}';
+    },
     extractExpected: (tc) => JSON.stringify(tc.expected_card),
     extractInput: (tc) => JSON.stringify(tc.input_data),
 }).run(process.argv.includes('--run-one'));