Documentation Index
Fetch the complete documentation index at: https://docs.minns.ai/llms.txt
Use this file to discover all available pages before exploring further.
The event_type field on every event uses a tagged enum. Exactly one variant must be provided.
Action
Use when the agent does something — calls an API, books a seat, sends an email.
{
"Action": {
"action_name": "book_movie_ticket",
"parameters": { "movie": "Interstellar", "seat": "H12" },
"outcome": { "Success": { "result": { "confirmation_id": "CONF-123" } } },
"duration_ns": 1500000
}
}
| Field | Type | Required | Description |
|---|
action_name | String | ✅ Yes | Name of the action executed |
parameters | JSON | ✅ Yes | Action parameters (any JSON, can be null) |
outcome | ActionOutcome | ✅ Yes | Result of the action |
duration_ns | u64 | ✅ Yes | Execution duration in nanoseconds |
ActionOutcome variants
{ "Success": { "result": { "confirmation_id": "CONF-123" } } }
{ "Failure": { "error": "Seat unavailable", "error_code": 409 } }
{ "Partial": { "result": { "booked": true }, "issues": ["Payment pending"] } }
Observation
Use when the agent receives external data that isn’t a direct message — price updates, sensor readings, API responses.
{
"Observation": {
"observation_type": "ticket_availability",
"data": { "remaining_seats": 5 },
"confidence": 1.0,
"source": "Theater-API"
}
}
| Field | Type | Required | Range | Description |
|---|
observation_type | String | ✅ Yes | — | Type of observation |
data | JSON | ✅ Yes | — | Observed data (any JSON) |
confidence | f32 | ✅ Yes | 0.0–1.0 | Confidence in observation |
source | String | ✅ Yes | — | Data source identifier |
Cognitive
Use to log the agent’s internal “thinking” process. This is crucial for strategy extraction.
{
"Cognitive": {
"process_type": "Reasoning",
"input": { "user_query": "I want a quiet seat" },
"output": { "decision": "Select back row" },
"reasoning_trace": [
"User requested quiet environment",
"Back rows have less foot traffic",
"Checking availability for row Z"
]
}
}
| Field | Type | Required | Description |
|---|
process_type | CognitiveType | ✅ Yes | Type of cognitive process |
input | JSON | ✅ Yes | Input to cognitive process |
output | JSON | ✅ Yes | Output from process |
reasoning_trace | String[] | ✅ Yes | Reasoning steps (can be []) |
CognitiveType values
| Value | Description |
|---|
"GoalFormation" | Forming a new goal |
"Planning" | Creating a plan |
"Reasoning" | General reasoning |
"MemoryRetrieval" | Retrieving from memory |
"LearningUpdate" | Updating learned knowledge |
Communication
Use for every message between the user and the agent.
{
"Communication": {
"message_type": "user_message",
"sender": 0,
"recipient": 1,
"content": { "text": "I'd like to book a ticket for tonight." }
}
}
| Field | Type | Required | Description |
|---|
message_type | String | ✅ Yes | Type of message (e.g., "user_message") |
sender | u64 | ✅ Yes | Sender agent ID |
recipient | u64 | ✅ Yes | Recipient agent ID |
content | JSON | ✅ Yes | Message content (any JSON) |
Learning
Use to log memory and strategy lifecycle events.
{
"Learning": {
"event": { "MemoryRetrieved": { "query_id": "q-123", "memory_ids": [1, 2, 3] } }
}
}
LearningEvent variants
| Variant | Fields | Description |
|---|
MemoryRetrieved | query_id: String, memory_ids: u64[] | Memories retrieved for a query |
MemoryUsed | query_id: String, memory_id: u64 | A specific memory was used |
StrategyServed | query_id: String, strategy_ids: u64[] | Strategies served for a query |
StrategyUsed | query_id: String, strategy_id: u64 | A specific strategy was used |
Outcome | query_id: String, success: bool | Final outcome of the learning cycle |
Context
Use for raw text content intended for claim extraction.
{
"Context": {
"text": "The user mentioned they prefer IMAX screenings and aisle seats.",
"context_type": "conversation",
"language": "en"
}
}
| Field | Type | Required | Description |
|---|
text | String | ✅ Yes | Raw text content |
context_type | String | ✅ Yes | Type: "conversation", "document", "transcript" |
language | String | null | No | Language hint (e.g., "en", "es") |
