---
title: "Ranked feed of public conversations where people seek alternatives to or are switching from a competitor."
method: GET
path: "/prism/leads"
tags: ["prism"]
---

# Ranked feed of public conversations where people seek alternatives to or are switching from a competitor.

`GET /prism/leads`

## Request

- Base URL: `https://www.socialcrawl.dev/v1`
- URL: `https://www.socialcrawl.dev/v1/prism/leads`
- Auth: API key in header `x-api-key`

Runs three exact-intent phrases ('alternative to X', 'X vs', 'switching from X') through content_analysis search in parallel, plus a Reddit (sort=new) and a Hacker News leg, then dedupes them into one feed. Since methodology leads/2 the ranking is judged: an AI classifier reads each candidate and answers five narrow questions (what the author is doing, whether the row is about the competitor product rather than a namesake, whether the author is a vendor, whether your offer would help them, how urgent the need is), and code combines those answers with recency and engagement into lead_score = (0.35 x P(asking, comparing or switching) + 0.20 x P(fits offer) + 0.15 x urgency/3 + 0.15 x recency + 0.15 x engagement) x P(about) x (1 - P(vendor)). Each lead carries intent { label, probabilities, about, seller, fits_offer, urgency, confidence } and lead_score; the label reads 'uncertain' when the classifier has no clear winner. Rows with P(about) below 0.3 (the snowflake on a Christmas jumper) are removed and their URLs listed in dropped_ids. The keyword intent_score and phrase_matched from leads-v1 are still on every row, and if the classifier is unavailable the feed falls back to that keyword order with _warnings: ['intent_unavailable'], intent: null and lead_score: null ('intent_partial' when only some rows could be judged). Conversation-level intent only: page-level matching, no author PII harvesting (web leads carry no handle by design, and no handle is sent to the classifier). Coverage spine is the three CA legs and does not depend on the classifier: <0.5 succeed → 50% refund; all five legs fail → full refund. Registers cleanly for a recurring-lead-feed monitor.

## Query parameters

- `competitor` string, required
- `product_category` string
- `freshness` string
- `include` string
- `limit` integer
- `offer` string

## Headers

- `Cache-Control` string
- `Idempotency-Key` string

## Response `200`

Successful response

- object
  - `success` boolean, required: Whether the request succeeded
  - `platform` string, required: Platform name
  - `endpoint` string, required: API endpoint path
  - `data` object, required: Platform-specific response data
    - `metrics` object: Key performance metrics
      - `total_views` integer: Total view count
      - `total_likes` integer: Total like count
      - `total_comments` integer: Total comment count
      - `engagement_rate` number: Computed engagement rate (0-1)
    - `period` string, nullable: Time period for the analytics data
    - `breakdown` object[], nullable: Per-item or per-period breakdown
    - `_warnings` string[]: Non-fatal notices about this response (field-map drift, clamped computed values). Advisory only: its presence never means the request failed. Omitted entirely when there is nothing to report, so treat absent as 'no warnings'.
  - `credits_used` integer, required: Number of credits consumed
  - `credits_remaining` integer, nullable, required: Current account balance, or null when this response could not establish one. Free cache hits always return null because they do not establish an authoritative post-request balance. An idempotency replay also returns null when its current-balance lookup fails. It is never a fabricated zero, so a zero here means the account really is empty. The authoritative reading is the one on a billed response, which comes straight from the deduction.
  - `request_id` string, required: Unique request identifier for support
  - `cached` boolean, required: Whether the response was served from cache
  - `idempotent_replay` boolean: True only when this response is an idempotency replay
  - `meta` object: Optional. Present only when this call completed a pattern that one other call does more cheaply or in one step, at most once per pattern per API key per day. Send the header `X-SocialCrawl-Hints: off` to never receive it.
    - `hint` object: One suggestion: what to call instead and why, with the price when the suggested call costs more than the calls it replaces.
      - `code` string: creator_card | batch_profiles | search_multi | incremental_sync | comment_labels | search_repeat
      - `message` string: One sentence for a person or an agent to read.
      - `method` string: GET or POST.
      - `path` string: The call the hint points to.

## Other responses

- `400`: Invalid request - missing or malformed parameters
- `401`: Unauthorized - missing or invalid API key
- `402`: Payment required - the account balance is too low, or the calling key has spent its own per-key credit limit
- `404`: Not found - the endpoint does not exist, or the requested resource was not found upstream
- `405`: Method not allowed - wrong HTTP verb for this endpoint
- `409`: Conflict - an idempotent request is still in flight, its completed body exceeded the replay limit, or the cohort resource is not in a state that allows this operation
- `410`: Gone - the discover run's retention passed and its rows were deleted; start a new run
- `413`: Payload too large: the JSON request body exceeds the 1 MB size limit and is rejected before parsing, or a single cohort result cannot fit beneath the 1 MB response-page ceiling
- `422`: Idempotency payload mismatch - this Idempotency-Key was already used with a different request payload; or a discover request with async=false is over what one call can answer (NEEDS_ASYNC)
- `429`: Rate or concurrency limit exceeded. `RATE_LIMITED`: more than 600 requests in a 1-minute sliding window on this API key (headers `X-RateLimit-Limit`/`Remaining`/`Reset`; `Retry-After` is seconds until the window resets). `CONCURRENCY_LIMIT`: more than 50 simultaneous in-flight requests (headers `X-Concurrency-Limit`/`Remaining`; short static `Retry-After`). Both are unbilled. Honor `Retry-After`, then back off with jitter. See /docs/rate-limits.
- `500`: Internal server error - an immediate refund is attempted; credits_used reports the committed net charge and exact compensation is queued if that refund cannot commit
- `502`: Upstream error - the platform returned an error; an immediate refund is attempted and credits_used reports the committed net charge
- `503`: Service unavailable - the request is unbilled when rejected before deduction, otherwise an immediate refund is attempted and credits_used reports the committed net charge
- `504`: Gateway timeout - the request was still running close to the platform time limit and is answered as UPSTREAM_ERROR, the same code as 502. An immediate refund is attempted and credits_used reports the committed net charge. When details.retry_with is set, retryable is false and there is no Retry-After

---

[API](https://skmtc.dev/socialcrawl/apis/socialcrawl-api.md) · [All operations](https://skmtc.dev/socialcrawl/apis/socialcrawl-api/llms.txt) · [OpenAPI document](https://skmtc.dev/socialcrawl/apis/socialcrawl-api/revisions/a0ad2511004a?raw)