Common Workflows

You: Create a new customer support bot called "HelpDesk" in workspace 5

AI: I'll create that app for you.
[Calls create_app with name="HelpDesk", workspaceId=5]
Created app "HelpDesk" with ID 847.

You: Set up the system prompt to be a helpful support agent that answers questions
about our product and escalates complex issues

AI: I'll update the system prompt.
[Calls update_system_prompt]
Updated system prompt for HelpDesk.

You: Add our documentation site as a knowledge source: https://docs.example.com

AI: I'll add that URL as a knowledge source with default crawl depth.
[Calls add_url_source with url="https://docs.example.com"]
Added knowledge source. Status: PROCESSING.
It will be ready for use in a few minutes.

You: Test it with "What is your refund policy?"

AI: I'll send that test message.
[Calls chat_with_app with message="What is your refund policy?"]

Bot response: "Based on our documentation, our refund policy allows..."

You: List all my apps

AI: [Calls list_apps]
Found 12 apps:
- ID 123: FAQ Bot
- ID 124: Sales Assistant
- ID 125: Onboarding Guide
...

You: Update all apps that use GPT-3.5 to use GPT-4o-mini instead

AI: I'll check each app's model and update the ones using GPT-3.5.
[Calls get_app for each, then update_model where needed]
Updated 7 apps from GPT-3.5 to GPT-4o-mini:
- FAQ Bot (123)
- Sales Assistant (124)
- Product Helper (128)
...

You: Update all 50 of my apps to use the new system prompt template,
but space out the requests to avoid rate limiting

AI: I'll update apps with a 2-second delay between each call
to stay within the BUILDER tier limit of 30 requests/minute.
[Updates apps one by one with delays]
Progress: 25/50 apps updated...

You: List knowledge sources in my old FAQ app (ID 100)

AI: [Calls list_knowledge_sources]
Found 8 knowledge sources:
1. docs.example.com (URL, 45 pages)
2. Product Manual.pdf (FILE)
3. FAQ-2024.docx (FILE)
...

You: Add all the same URLs to my new FAQ app (ID 200)

AI: I'll add each URL source to the new app.
[Calls add_url_source for each URL]
Added 5 URL sources to app 200:
- docs.example.com
- help.example.com
- blog.example.com/faq
...

Note: File sources need to be re-uploaded manually since I can't
access the original file content.

You: Show me analytics for my main support bot (ID 123) for the past week

AI: [Calls get_app_analytics with since="7 days ago"]

Analytics for Support Bot (last 7 days):
- 1,247 conversations
- 3,891 messages
- 312 unique users
- 4.2/5.0 average satisfaction
- 89% resolution rate

You: What are the most common questions?

AI: [Calls get_top_queries]
Top queries:
1. "How do I reset my password?" (142 times)
2. "What are your business hours?" (98 times)
3. "Can I get a refund?" (87 times)
4. "How do I contact support?" (76 times)
...

You: Find conversations where users mentioned "frustrated" or "angry"

AI: [Calls search_conversations with query="frustrated angry"]
Found 23 conversations with negative sentiment:
- Session abc123: User frustrated about billing...
- Session def456: User angry about wait time...

You: Create an evaluation for my FAQ bot based on a recent good conversation

AI: First, let me find a recent successful conversation.
[Calls search_conversations]
Found a conversation with clear question/answer pairs from yesterday.

[Calls get_conversation to read the full messages]
I'll extract test cases from those exchanges and create an eval.

[Calls create_eval with name="FAQ Quality Test" and 8 test cases]
Created evaluation "FAQ Quality Test" with 8 test cases. Would you like to review them?

You: Run the evaluation

AI: [Calls run_eval]

Evaluation complete. Results:
- Overall: PASSED (7/8 tests)
- Pass rate: 87.5%

Failing test:
- Test 7: Complex billing scenario
  Expected: response about prorated billing
  Got: generic "contact support" response

You: Update the system prompt to better handle billing questions,
then run the eval again

AI: [Calls improve_system_prompt with focus="billing questions"]
[Calls run_eval]
All 8 tests now pass!

You: Create a tag called "pricing_intent" for my sales bot to detect
when users ask about pricing

AI: [Calls create_tag]
Created tag "pricing_intent". Would you like me to generate
training utterances for it?

You: Yes, generate 200 utterances for pricing questions in a B2B
software context

AI: The tag is now trained with utterances. Examples:
- "How much does your enterprise plan cost?"
- "What's the pricing for 100 users?"
- "Do you offer volume discounts?"
- "Is there a free trial available?"
...

You: Show me the last 10 messages that matched this tag

AI: [Calls list_tag_instances]
Recent pricing intent matches:
1. "What's your pricing model?" - 2 hours ago
2. "How much would it cost for my team of 50?" - 5 hours ago
3. "Do you have monthly billing options?" - yesterday
...

You: Show me all deployments for my support bot

AI: [Calls list_slack_deployments, get_email_deployment,
get_whatsapp_deployment, get_voice_config]

Support Bot Deployments:
- Slack: Connected to "Acme Corp" workspace
  - Channels: #support, #sales
  - Last message: 2 hours ago

- Email: Enabled
  - Address: support@acme.chipp.ai
  - Last email: yesterday

- WhatsApp: Not configured

- Voice: Enabled
  - Phone: +1 (555) 123-4567
  - Language: English
  - Last call: 3 days ago

You: Update the voice config to also support Spanish

AI: [Calls update_voice_config with language="es,en"]
Updated voice configuration. The bot now supports both
English and Spanish phone calls.

You: My FAQ bot sometimes gives overly long responses.
Help me improve the prompt to be more concise.

AI: Let me analyze your current prompt and suggest improvements.
[Calls get_system_prompt]
[Calls improve_system_prompt with focus="conciseness"]

I've updated the prompt with these changes:
1. Added instruction to keep responses under 3 sentences when possible
2. Added "Get to the point quickly" guidance
3. Removed redundant "be helpful" phrasing

Would you like to test the changes?

You: Yes, test with "What is your return policy?"

AI: [Calls chat_with_app]

Before: "Thank you for your question about our return policy!
We're happy to help you understand how returns work. Our return
policy allows you to return any item within 30 days of purchase
for a full refund. The item must be in its original condition
with all tags attached. Would you like me to explain the
return process in detail?"

After: "You can return items within 30 days for a full refund.
Items must be in original condition with tags. Need help
starting a return?"

Much more concise!

You: Give me a summary of all my apps' performance this week

AI: [Calls list_apps]
[Calls get_app_analytics with since="7 days ago" for each app]

Weekly Performance Summary (12 apps):

Top Performers:
1. Support Bot: 1,247 conversations, 4.5 stars
2. Sales Assistant: 432 conversations, 4.3 stars
3. FAQ Bot: 298 conversations, 4.1 stars

Needs Attention:
- Onboarding Guide: 2.8 stars (down from 3.5 stars)
- Product Helper: Only 12 conversations (was 89 last week)

Total across all apps:
- 2,891 conversations
- 8,234 messages
- 892 unique users

You: What's wrong with the Onboarding Guide?

AI: [Calls search_conversations for the Onboarding Guide app]
[Calls get_top_queries]

Found the issue: Users are asking about the new "Teams" feature
that was added last week, but the bot doesn't have knowledge about it.

Top unanswered queries:
1. "How do I create a team?" (34 times)
2. "What are team permissions?" (28 times)
3. "Can I invite external users to my team?" (19 times)

Recommendation: Add your Teams documentation as a knowledge source.

"List all my apps"
"Create an app called [name] in workspace [id]"
"Delete app [id]"
"What model is app [id] using?"

"Add [url] to app [id]'s knowledge base"
"List knowledge sources for app [id]"
"Refresh all knowledge sources for app [id]"

"Test app [id] with message '[message]'"
"Run all evaluations for app [id]"
"Create an evaluation from conversation [session_id]"

"Show analytics for app [id] for the past week"
"Find conversations mentioning '[keyword]'"
"What are the top questions users ask app [id]?"

"Create a tag called '[name]' for app [id]"
"Show messages that matched tag '[name]'"