AI Assistant¶
The AI Assistant is an intelligent, context-aware chat interface that helps you get instant answers about your POCs, tasks, resources, and project data. Powered by LLMs and integrated with your existing POC data, it streamlines information retrieval and decision-making.
Beta Feature
The AI Assistant is currently in Beta. Features and APIs may change. We welcome your feedback!
Overview¶
The AI Assistant provides:
- Natural language queries about POCs, tasks, and resources
- Instant data retrieval across all your POC information
- Context-aware responses based on your role and access permissions
- Tool integration with POC Manager APIs through MCP (Model Context Protocol)
- Secure sessions with automatic timeouts and role-based access control
Getting Started¶
Prerequisites¶
Before using the AI Assistant, your Tenant Admin must:
- Enable AI Assistant in tenant settings
- Configure the API key for secure communication
Cloud-Hosted Models
The AI Assistant uses cloud-hosted language models through an Ollama-compatible API. Your Tenant Admin manages the API key configuration. No local installation is required.
Getting an Ollama API Key¶
Step 1: Create an Ollama Account¶
- Visit ollama.com
- Click Sign Up in the top-right corner
- Enter your email address and create a password
- Verify your email by clicking the link sent to your inbox
- Complete your profile information
Step 2: Generate an API Key¶
- Log in to your Ollama account
- Navigate to Account Settings → API Keys
- Click Create New API Key
- Give it a descriptive name (e.g., "POC Manager")
- Select the appropriate permissions for the AI Assistant
- Click Generate
- Copy the API key immediately—you won't be able to see it again
Secure Your API Key
- Store it in a password manager or secure secrets vault
- Never commit it to version control
- Never share it with team members directly
- Regenerate it if you suspect it's been compromised
Step 3: Add to POC Manager¶
Once you have your API key, follow the Enabling AI Assistant instructions above to configure it in your tenant settings.
Opening the AI Assistant¶
- Look for the AI Assistant button (💬 with "AI Assistant" label) in the bottom-right corner of the screen
- Click to open the chat interface
- You'll see the AI Assistant status:
- ✅ Ready: AI Assistant is enabled and configured
- ❌ Not Enabled: Contact your Tenant Admin to enable it
- ⚠️ Not Configured: API key is not set (Tenant Admin action required)
Starting a Conversation¶
- Click in the message input field at the bottom of the chat window
- Type your question or request
- Press Enter or click Send to submit
- The AI Assistant will process your query and provide a response
The conversation history is maintained in your current session (10-minute inactive timeout).
Available to User Roles¶
| Role | Access |
|---|---|
| Platform Admin | ❌ Not available |
| Tenant Admin | ✅ Yes |
| Administrator | ✅ Yes |
| Sales Engineer | ✅ Yes |
| Customer | ❌ Not available |
Note
Customers and Platform Admins cannot access the AI Assistant. The feature is designed for internal team members who manage POCs.
Enabling AI Assistant¶
For Tenant Admins¶
To enable the AI Assistant for your tenant:
- Navigate to Tenant Settings (admin only)
- Scroll to AI Assistant section
- Click the Enable AI Assistant toggle
- A warning dialog will appear. Review the requirements:
- You have a valid API key for the cloud service
- You understand that chat messages are processed by the cloud service
- Click I Understand, Enable to confirm
- Enter your API Key in the provided field
- Leave blank if updating an existing key
- Click Save AI Assistant Settings
Cloud-Hosted Models
The AI Assistant uses cloud-hosted language models. Your Tenant Admin manages the API key configuration for secure access. No local installation or infrastructure is required.
Using the AI Assistant¶
Example Queries¶
Here are common types of questions you can ask:
POC Information¶
"What POCs are currently in progress?"
"Show me all POCs for Acme Corp"
"How many customers have completed their POCs?"
Task Management¶
"What tasks are assigned to me?"
"Which tasks in the Marketing POC are blocked?"
"List all incomplete tasks from this quarter"
Resource Discovery¶
"Find all code snippets related to authentication"
"Show me resources linked to API integration tasks"
"What documentation do we have for the payment module?"
Progress & Metrics¶
"What's the overall completion rate for active POCs?"
"Which POCs are at risk of missing their end date?"
"How many participants does the Analytics POC have?"
Participant Information¶
"Who are the participants in the CRM POC?"
"List all sales engineers assigned to active POCs"
"What customers are currently engaged?"
Chat Features¶
Markdown Support¶
Responses are rendered with full markdown support, including: - Bold text and italic text - Code snippets with syntax highlighting - Bullet points and numbered lists - Links and tables - Quotes and blockquotes
Copy Responses¶
Each response includes a Copy button to quickly copy the content to your clipboard.
Session Management¶
- Conversation history is maintained within a session
- Sessions auto-timeout after 10 minutes of inactivity
- Click New Chat to start a fresh conversation
- Your session is private to your user account
Message History¶
The AI Assistant remembers recent messages in your conversation to provide contextual responses. Use this to build on previous queries:
User: "How many active POCs do we have?"
AI: [responds with list]
User: "Who are the participants in the first one?"
AI: [responds with participants from first POC]
How It Works¶
Architecture¶
The AI Assistant combines several technologies:
┌─────────────────┐
│ Chat Interface │ (Your browser)
│ (React Component)│
└────────┬────────┘
│
│ HTTP/REST
▼
┌─────────────────────────┐
│ POC Manager Backend │
│ - Session Management │
│ - Role-based Access │
│ - Security/JWT Auth │
└────────┬────────────────┘
│
├─────────────┬──────────────┐
▼ ▼ ▼
┌────────┐ ┌──────────┐ ┌─────────┐
│ Ollama │ │ MCP │ │Database │
│ LLM │ │ Tools │ │Content │
└────────┘ └──────────┘ └─────────┘
Behind the Scenes¶
- You send a message to the chat interface
- Session validated - your message is added to conversation history
- LLM receives prompt - your question + system instructions + tools catalog
- Tool execution - if needed, the LLM calls MCP tools to fetch real data
- MCP tools use your auth token to ensure RBAC
- Only data you can access is returned
- Response generated - the LLM synthesizes a natural language response
- Response sent back - rendered in chat with markdown formatting
Role-Based Access Control¶
The AI Assistant respects your role-based permissions:
- You can only query data you have access to
- MCP tools verify your roles before returning data
- Responses only include POCs, tasks, and resources in your tenant
- Tenant data is completely isolated
Conversation Tips¶
Be Specific¶
✅ Good: "What are the blocked tasks in the 'API Integration' POC?"
❌ Poor: "Show me tasks"
Use Context¶
✅ Good: "How many participants does the Acme Corp POC have?"
❌ Poor: "participants?" (unclear which POC)
Break Down Complex Queries¶
✅ Better:
1. "List all POCs with status = 'In Progress'"
2. "How many tasks does the first POC have?"
3. "Show me any blocked tasks"
❌ Complex:
"Show me all in-progress POCs with their task counts and blocked tasks"
Don't Ask for Sensitive Data¶
❌ "What is the admin password?"
❌ "Show me all user email addresses"
The AI Assistant won't share sensitive information outside your access level.
Session Management¶
Session Timeout¶
- Timeout Duration: 10 minutes of inactivity
- What happens: Automatic logout, chat history cleared
- Reset: Start a new chat
Manual Controls¶
| Action | Effect |
|---|---|
| Close Chat | Session remains active in background |
| New Chat | Creates new session, clears message history |
| Page Refresh | Session persists (within timeout) |
| Logout | All sessions end |
| Timeout | Toast notification, automatic closure |
Using Multiple Sessions¶
You can open the AI Assistant in multiple browser tabs/windows. Each session: - Is independent - Has its own 10-minute timeout - Maintains separate conversation history - Uses your current user credentials
Configuration (Tenant Admin)¶
Ollama API Key Management¶
Important
- API keys are encrypted before storage
- Keys are only used for AI Assistant sessions
- Never share your API key with others
- Store keys in a secure secrets manager
Updating Settings¶
To modify AI Assistant settings:
- Go to Tenant Settings → AI Assistant
- Make changes:
- Toggle AI Assistant on/off
- Update Ollama API key (leave blank to keep existing)
- Change any configuration options
- Click Save AI Assistant Settings
- See confirmation message
Disabling AI Assistant¶
To turn off the feature:
- In Tenant Settings → AI Assistant
- Click the Enable AI Assistant toggle to OFF
- Click Save AI Assistant Settings
- Users will see "AI Assistant is not enabled" message
All existing sessions will end, and the button will disappear from the interface.
MCP Tools¶
The AI Assistant can use MCP Tools to query POC data in real-time. These tools are automatically discovered from the POC Manager backend.
Available Tools (Example)¶
When enabled, the AI Assistant has access to tools like:
list_pocs- Get information about POCslist_poc_tasks- Retrieve tasks for a specific POClist_participants- Find POC participantslist_users- Look up user informationget_poc- Get detailed POC information
Note
Tools available depend on your Tenant Admin configuration (via MCP_TOOLS environment variable).
What the LLM Does with Tools¶
The LLM automatically decides whether to use tools:
User Query: "What tasks are in the Auth POC?"
LLM Decision Flow:
1. Parse query: "Find tasks in Auth POC"
2. Determine tools needed: list_pocs (to find Auth POC), list_poc_tasks (get tasks)
3. Call tools with appropriate parameters
4. Receive structured data
5. Format response in natural language:
"The Auth POC has 5 tasks: Authentication..."
You never see the technical details—just the natural language response!
Limitations¶
| Limitation | Details |
|---|---|
| No Data Modification | The AI Assistant is read-only. It cannot create, update, or delete POCs, tasks, etc. Make changes directly in POC Manager. |
| Model Availability | Responses depend on your Ollama service being available and responsive. |
| Knowledge Cutoff | The LLM has a training cutoff date and may not know about recent features. |
| Context Limits | Long conversations may be summarized to stay within token limits. |
| Hallucination Risk | While MCP tools help, the LLM may occasionally provide inaccurate information if tools don't fully answer the question. |
| No Real-Time Events | The AI Assistant reflects current database state; it's not aware of live updates from other users. |
Troubleshooting¶
"AI Assistant is not enabled"¶
Solution: Contact your Tenant Admin and ask them to enable the AI Assistant in Tenant Settings.
"AI Assistant is enabled but Ollama API key is not configured"¶
Solution: Your Tenant Admin needs to add the Ollama API key in Tenant Settings → AI Assistant.
Response is taking too long¶
Solution: 1. Check if your Ollama service is running 2. Check network connectivity to Ollama endpoint 3. Try a simpler query first 4. Wait up to 2 minutes (default timeout)
"I'm having trouble processing your request"¶
Possible causes: - Ollama service is down - API key is incorrect or expired - Network connectivity issue - LLM model not available on Ollama service
Solution:
1. Notify your Tenant Admin
2. Verify Ollama service is running: curl https://OLLAMA_URL/api/models
3. Verify API key is correct
4. Try again in a few moments
Chat messages are not appearing¶
Solution: 1. Check your internet connection 2. Verify browser console for errors (press F12) 3. Try closing and reopening the AI Assistant 4. Refresh the page 5. Clear browser cache if issues persist
Session timed out unexpectedly¶
Solution: 1. This is normal after 10 minutes of inactivity 2. Click New Chat to start a fresh conversation 3. If you need longer sessions, keep the chat window active
Best Practices¶
✅ Do¶
- Start simple - Ask one thing at a time
- Be specific - Include POC names, dates, or other identifiers
- Verify results - Check responses against POC Manager for accuracy
- Use feedback - Report issues to improve the AI Assistant
- Keep sessions short - Start new chats for different topics
❌ Don't¶
- Ask for sensitive data - Passwords, API keys, etc. (you shouldn't have access anyway)
- Rely entirely on AI - Use POC Manager as the source of truth
- Leave sensitive responses visible - Copy-paste responses to others carefully
- Share API keys - Never include your Ollama API key in conversations
FAQ¶
Q: Who can see my conversations? A: Only you can see your chat history. Sessions are user-specific and encrypted in transit. Your Tenant Admin can see that the feature is being used but not conversation content.
Q: Can the AI Assistant modify my POC data? A: No, it's read-only. It can only retrieve and display information. Make changes directly in POC Manager.
Q: Is my data sent to OpenAI or other external services? A: No, your data is sent only to your Ollama API endpoint, which you control. It's a bring-your-own-model feature.
Q: Can I use any LLM with the AI Assistant? A: Yes, as long as it's compatible with Ollama. Your Tenant Admin configures the model in settings.
Q: How much does the AI Assistant cost? A: Costs depend on your Ollama setup. If you self-host, it's free (hardware costs only). If using a cloud provider, pricing varies by provider and usage.
Q: What happens to my chat history? A: History is stored in your session during the 10-minute window. After timeout, it's automatically deleted. We do not retain chat history between sessions.
Q: Can I export chat history? A: Currently, no built-in export. You can manually copy responses using the Copy button.
Q: What's the maximum message length? A: The AI Assistant handles typical messages (< 4,000 characters). Extremely long messages may be truncated.
Support¶
If you encounter issues or have feedback:
- Check this documentation - Your question may be answered here
- Contact your Tenant Admin - They can verify settings and enable features
- Report bugs - Include what you asked and what response you got
- Request features - Suggest what would make the AI Assistant more helpful
Last Updated: February 2024
Status: Beta
Feedback: We'd love to hear from you! This feature is actively being improved based on user input.