Spark Suggestions

Spark produces one focused AI suggestion per request. It resolves context from your Chainies, Bits, personalization memory, and active sessions, then generates a concise suggestion without starting a full conversation session.

Use Spark for lightweight in-app moments: "what should I work on next?", "turn this comment into an action", "suggest a Bit for this Chainy".

Endpoint

Method	Path	Description	Auth	Rate Limit
POST	`/ai/spark/suggestions`	Create a context-aware suggestion	JWT + Spark entitlement	30/min

POST /ai/spark/suggestions

Authentication: JWT Bearer token required.
Entitlement: Active Spark entitlement required. Available on Plus and above (not included in the Free plan).
Rate limit: 30 requests per minute.
Idempotency: Pass an idempotency-key header to deduplicate retries — identical keys return the cached result without re-invoking the LLM.

Request

Request Headers

Header	Required	Description
`Authorization`	Yes	`Bearer <token>`
`Content-Type`	Yes	`application/json`
`idempotency-key`	No	Unique key to deduplicate the request

Request Body

All fields are optional. If no fields are provided the server uses the auto_suggest mode.

Field	Type	Constraints	Description
`mode`	string	`auto_suggest`, `prompt_suggest`, `next_bit`, `comment_result`	Explicit mode. Omit to let the server resolve automatically.
`prompt`	string	max 4000 chars	Primary instruction or question.
`comment`	string	max 4000 chars	Raw comment to transform into a suggestion.
`chainyIds`	string[]	max 10, each a valid UUID	Ground the suggestion in specific Chainies.
`bitIds`	string[]	max 25, each a valid UUID	Ground the suggestion in specific Bits.
`contextOptions.useMemory`	boolean	—	Set `false` to suppress memory context for this request.
`contextOptions.useLatestSessions`	boolean	—	Set `false` to suppress recent-sessions context for this request.

Mode auto-resolution (when mode is omitted):

comment provided → comment_result
prompt provided → prompt_suggest
chainyIds provided → next_bit
Nothing → auto_suggest

Response

Response Example

json

{
  "data": {
    "suggestionId": "cm9spark01",
    "mode": "next_bit",
    "title": "Add a pronunciation practice session",
    "content": "Your Spanish Chainy has strong vocabulary coverage but no pronunciation Bits. Adding a 10-minute daily audio session would round out the practice cycle.",
    "structuredResult": {
      "type": "next_bit",
      "payload": {
        "suggestedTitle": "Pronunciation: 10-min daily audio",
        "suggestedPriority": "high"
      }
    },
    "contextUsed": ["chainies", "bits", "memory"],
    "blockedContext": [
      {
        "category": "sessions",
        "reason": "sessions_disabled_by_preference"
      }
    ],
    "chainyIdsUsed": ["a1b2c3d4-e5f6-7890-abcd-ef1234567890"],
    "bitIdsUsed": [
      "b2c3d4e5-f6a7-8901-bcde-f12345678901",
      "c3d4e5f6-a7b8-9012-cdef-123456789012"
    ],
    "personaUsed": "coach",
    "provider": "google",
    "model": "gemini-2.5-flash",
    "createdAt": "2026-05-22T09:30:00.000Z"
  }
}

Response Fields

Field	Type	Description
`suggestionId`	string	Persisted suggestion ID
`mode`	string	Resolved mode used: `auto_suggest`, `prompt_suggest`, `next_bit`, `comment_result`
`title`	string	Short suggestion headline (max 200 chars)
`content`	string	Full suggestion text (max 2000 chars)
`structuredResult`	object \| null	Optional structured output (see below)
`contextUsed`	string[]	Context categories that fed the suggestion
`blockedContext`	object[]	Context categories that were suppressed and why
`chainyIdsUsed`	string[]	Chainy IDs that grounded the suggestion
`bitIdsUsed`	string[]	Bit IDs that grounded the suggestion
`personaUsed`	string \| null	AI persona key if a persona was applied
`provider`	string	AI provider used (e.g. `google`, `anthropic`)
`model`	string	Model identifier
`createdAt`	string	ISO 8601 timestamp

`structuredResult` object

Present when the LLM returns structured data in addition to the narrative text.

Field	Type	Description
`type`	string	`next_bit`, `idea`, `transformation`, or `generic`
`payload`	object	Type-specific fields (e.g. `suggestedTitle`, `suggestedPriority`)

`contextUsed` values

Value	Meaning
`prompt`	User's `prompt` field was used
`comment`	User's `comment` field was used
`chainies`	Chainy data was retrieved
`bits`	Bit data was retrieved
`memory`	Personalization memory was used
`sessions`	Recent AI sessions were used
`preferences`	AI persona or language preference was applied

`blockedContext` object

Field	Type	Description
`category`	string	The `contextUsed` category that was blocked
`reason`	string	Why it was blocked (see table below)

Reason	Cause
`memory_disabled_by_preference`	Memory disabled in your AI preferences
`memory_disabled_by_client`	Request sent `contextOptions.useMemory: false`
`sessions_disabled_by_preference`	Session context disabled in your AI preferences
`sessions_disabled_by_client`	Request sent `contextOptions.useLatestSessions: false`
`empty_input`	No prompt, comment, or Chainy IDs were provided
`no_bits_available`	The requested Chainies exist but contain no Bits

Error Responses

Status	Code	When
400	`VALIDATION_FAILED`	Mode/input mismatch (e.g. `next_bit` without `chainyIds`)
403	`FORBIDDEN`	`chainyIds` or `bitIds` do not belong to your account
429	`RATE_LIMIT_EXCEEDED`	More than 30 requests per minute
502	`BAD_GATEWAY`	AI provider returned an empty or failed response
503	`SERVICE_UNAVAILABLE`	No AI provider available

Code Examples

Use the id of one of your Chainies as $CHAINY_ID in the "next bit" examples below.

bash

curl -X POST "$BASE_URL/ai/spark/suggestions" \
  -H "Authorization: Bearer $TOKEN" \
  -H "Content-Type: application/json" \
  -d '{}'

bash

curl -X POST "$BASE_URL/ai/spark/suggestions" \
  -H "Authorization: Bearer $TOKEN" \
  -H "Content-Type: application/json" \
  -H "idempotency-key: req_abc123" \
  -d '{
    "mode": "prompt_suggest",
    "prompt": "What is the highest priority thing I should tackle today?"
  }'

bash

curl -X POST "$BASE_URL/ai/spark/suggestions" \
  -H "Authorization: Bearer $TOKEN" \
  -H "Content-Type: application/json" \
  -d '{
    "mode": "next_bit",
    "chainyIds": ["'"$CHAINY_ID"'"]
  }'

bash

curl -X POST "$BASE_URL/ai/spark/suggestions" \
  -H "Authorization: Bearer $TOKEN" \
  -H "Content-Type: application/json" \
  -d '{
    "mode": "comment_result",
    "comment": "I keep losing focus after lunch and end up skipping my language practice."
  }'

javascript

// Auto suggest
const autoResponse = await fetch(`${BASE_URL}/ai/spark/suggestions`, {
  method: "POST",
  headers: {
    Authorization: `Bearer ${TOKEN}`,
    "Content-Type": "application/json",
  },
  body: JSON.stringify({}),
});
const { data: autoSuggestion } = await autoResponse.json();

// Prompt mode with idempotency
const promptResponse = await fetch(`${BASE_URL}/ai/spark/suggestions`, {
  method: "POST",
  headers: {
    Authorization: `Bearer ${TOKEN}`,
    "Content-Type": "application/json",
    "idempotency-key": "req_abc123",
  },
  body: JSON.stringify({
    mode: "prompt_suggest",
    prompt: "What is the highest priority thing I should tackle today?",
  }),
});
const { data: promptSuggestion } = await promptResponse.json();

// Next bit for a Chainy
const chainyId = process.env.CHAINY_ID; // id of the Chainy to ground the suggestion in

const nextBitResponse = await fetch(`${BASE_URL}/ai/spark/suggestions`, {
  method: "POST",
  headers: {
    Authorization: `Bearer ${TOKEN}`,
    "Content-Type": "application/json",
  },
  body: JSON.stringify({
    mode: "next_bit",
    chainyIds: [chainyId],
  }),
});
const { data: nextBitSuggestion } = await nextBitResponse.json();

python

import os
import requests

BASE_URL = "https://api.chainabit.com/api/v1"
headers = {"Authorization": f"Bearer {TOKEN}", "Content-Type": "application/json"}

# Auto suggest
res = requests.post(f"{BASE_URL}/ai/spark/suggestions", headers=headers, json={})
data = res.json()["data"]

# Prompt mode
res = requests.post(
    f"{BASE_URL}/ai/spark/suggestions",
    headers={**headers, "idempotency-key": "req_abc123"},
    json={
        "mode": "prompt_suggest",
        "prompt": "What is the highest priority thing I should tackle today?",
    },
)
data = res.json()["data"]

# Next bit
chainy_id = os.environ["CHAINY_ID"]  # id of the Chainy to ground the suggestion in

res = requests.post(
    f"{BASE_URL}/ai/spark/suggestions",
    headers=headers,
    json={
        "mode": "next_bit",
        "chainyIds": [chainy_id],
    },
)
data = res.json()["data"]

Plan availability

Plan	Spark access	Monthly suggestion limit
Free	No	—
Plus	Yes	150
Pro	Yes	500
Pro (higher capacity)	Yes	Unlimited

Calling this endpoint on a plan without Spark access returns 403 FORBIDDEN.

Modes

`auto_suggest`

No inputs required. Spark reads your recent Bits, Chainies, memory, and sessions and generates an unprompted suggestion — useful for "surprise me" or "what's next?" moments.

`prompt_suggest`

Requires prompt. Generates a suggestion grounded in your explicit question or instruction combined with your available context.

`next_bit`

Requires at least one chainyId. Spark analyses the Chainy's existing Bits and suggests a concrete next Bit to create or action to take.

`comment_result`

Requires comment. Transforms a raw user comment (a note, thought, or observation) into an actionable suggestion.

Context controls

Grounding with specific IDs

Provide chainyIds and/or bitIds to pin the suggestion to specific content. Without IDs Spark pulls from your most recent Chainies and Bits automatically.

Suppressing context per request

Use contextOptions to override your account-level preferences for a single request:

json

{
  "contextOptions": {
    "useMemory": false,
    "useLatestSessions": false
  }
}

Account-level preferences

Memory and session context follow your AI preferences (ai_memory_enabled, use_latest_sessions). Manage these via the Preferences API.

Idempotency

Pass an idempotency-key header to make retries safe. If a suggestion was already created with the same key, the persisted result is returned immediately without calling the LLM again.

bash

curl -X POST "$BASE_URL/ai/spark/suggestions" \
  -H "Authorization: Bearer $TOKEN" \
  -H "Content-Type: application/json" \
  -H "idempotency-key: client-generated-unique-id" \
  -d '{ "prompt": "What should I focus on today?" }'

Keys are scoped to your account. Use a UUID or a hash of the request inputs as the key.

Spark Suggestions ​

Endpoint ​

POST /ai/spark/suggestions ​

Request ​

Request Headers ​

Request Body ​

Response ​

Response Example ​

Response Fields ​

structuredResult object ​

contextUsed values ​

blockedContext object ​

Error Responses ​

Code Examples ​

Plan availability ​

Modes ​

auto_suggest ​

prompt_suggest ​

next_bit ​

comment_result ​

Context controls ​

Grounding with specific IDs ​

Suppressing context per request ​

Account-level preferences ​

Idempotency ​

Spark Suggestions

Endpoint

POST /ai/spark/suggestions

Request

Request Headers

Request Body

Response

Response Example

Response Fields

`structuredResult` object

`contextUsed` values

`blockedContext` object

Error Responses

Code Examples

Plan availability

Modes

`auto_suggest`

`prompt_suggest`

`next_bit`

`comment_result`

Context controls

Grounding with specific IDs

Suppressing context per request

Account-level preferences

Idempotency