repull 0.2.3 → 0.2.4

data/lib/repull/models/webhook_event_catalog_entry.rb

@@ -1,7 +1,7 @@
 =begin
 #Repull API
 
-#The unified API for vacation rental tech. Connect to 50+ PMS platforms and 4 OTA channels through one REST API. Built-in AI operations for guest communication, pricing, and listing optimization.  ## Designed for AI agents Every error response on this API includes machine-parseable fields so an LLM (Claude in MCP, Cursor, Cline, GPT, etc.) can self-recover without escalating to a human: - `error.code` — stable string identifier (e.g. `invalid_params`, `rate_limit_exceeded`) - `error.message` — human-readable cause - `error.fix` — exact recovery steps (e.g. \"Pass `check_in_after` as ISO 8601: `?check_in_after=2026-01-15`\") - `error.docs_url` — link to the canonical write-up at `https://repull.dev/docs/errors/{code}` - `error.request_id` — id to correlate with server-side logs - `error.field` / `error.value_received` / `error.valid_values` / `error.did_you_mean` — when the error is parameter-specific - `error.retry_after` — seconds to wait before retrying (rate-limit + transient upstream)  `Access-Control-Expose-Headers` lists `x-request-id` and the `X-RateLimit-*` family so browsers can read them on cross-origin responses.  ## Quick Start 1. Get an API key at https://repull.dev/dashboard 2. Connect a PMS: `POST /v1/connect/{provider}` 3. List properties: `GET /v1/properties` 4. Get reservations: `GET /v1/reservations`  ## Authentication All requests require a Bearer token: ``` Authorization: Bearer sk_test_YOUR_API_KEY ```  Sandbox keys start with `sk_test_`, production with `sk_live_`.  ## Request Correlation (X-Request-ID) Every response carries an `X-Request-ID` header, e.g. `X-Request-ID: req_01HXY...`. Include this id in support tickets and bug reports — we can trace the full request lifecycle (auth, rate limit, handler, downstream calls, log row) from a single id.  You may set the header on the inbound request to forward your own trace id; we will echo it back instead of generating a new one. Accepted format: `^[\\\\w.-]{1,128}$`.  The id is also embedded in error envelopes as `request_id` so server-side log diffs work even when the response headers are stripped by an intermediate proxy.  ## Rate Limits The public API enforces a per-API-key sliding-window rate limit on top of the per-tier monthly + daily-AI quotas.  **Default policy:** 600 requests per 60 seconds, per API key. Sliding window — there is no fixed-minute boundary you can burst across.  Every response includes:  | Header | Meaning | |---|---| | `X-RateLimit-Limit` | Requests permitted in the current window. | | `X-RateLimit-Remaining` | Requests left in the current window after this call. | | `X-RateLimit-Reset` | Unix epoch (seconds) when the next slot opens. | | `X-RateLimit-Policy` | Machine-readable policy descriptor, e.g. `600;w=60`. | | `Retry-After` | Seconds to wait before retrying. **Only present on 429 responses.** |  **On 429 (rate_limit_exceeded):** the response body matches the standard error envelope with `code: \"rate_limit_exceeded\"`, plus `limit`, `window_seconds`, `retry_after`, and `request_id` fields. SDKs MUST honor `Retry-After` and use exponential backoff with jitter on subsequent retries — never a tight loop.  Recommended backoff: ``` sleep_ms = (Retry-After * 1000) + random(0..250) ```  Monthly + daily-AI tier quotas (`free`, `starter`, `custom`) are enforced separately and also surface as 429s; they include `tier`, `scope`, and `resets_at` fields.
+#The unified API for vacation rental tech. Connect to 50+ PMS platforms and 4 OTA channels through one REST API. Built-in AI operations for guest communication, pricing, and listing optimization.  ## Designed for AI agents Every error response on this API includes machine-parseable fields so an LLM (Claude in MCP, Cursor, Cline, GPT, etc.) can self-recover without escalating to a human: - `error.code` — stable string identifier (e.g. `invalid_params`, `rate_limit_exceeded`) - `error.message` — human-readable cause - `error.fix` — exact recovery steps (e.g. \"Pass `check_in_after` as ISO 8601: `?check_in_after=2026-01-15`\") - `error.docs_url` — link to the canonical write-up at `https://repull.dev/docs/errors/{code}` - `error.request_id` — id to correlate with server-side logs - `error.field` / `error.value_received` / `error.valid_values` / `error.did_you_mean` — when the error is parameter-specific - `error.retry_after` — seconds to wait before retrying (rate-limit + transient upstream)  `Access-Control-Expose-Headers` lists `x-request-id` and the `X-RateLimit-*` family so browsers can read them on cross-origin responses.  ## Quick Start 1. Get an API key at https://repull.dev/dashboard 2. Connect a PMS: `POST /v1/connect/{provider}` 3. List properties: `GET /v1/properties` 4. Get reservations: `GET /v1/reservations`  ## Authentication All requests require a Bearer token: ``` Authorization: Bearer sk_test_YOUR_API_KEY ```  Sandbox keys start with `sk_test_`, production with `sk_live_`.  ## Request Correlation (X-Request-ID) Every response carries an `X-Request-ID` header, e.g. `X-Request-ID: req_01HXY...`. Include this id in support tickets and bug reports — we can trace the full request lifecycle (auth, rate limit, handler, downstream calls, log row) from a single id.  You may set the header on the inbound request to forward your own trace id; we will echo it back instead of generating a new one. Accepted format: `^[\\\\w.-]{1,128}$`.  The id is also embedded in error envelopes as `request_id` so server-side log diffs work even when the response headers are stripped by an intermediate proxy.  ## Rate Limits The public API enforces a per-API-key sliding-window rate limit on top of the per-tier monthly + daily-AI quotas.  **Default policy:** 600 requests per 60 seconds, per API key. Sliding window — there is no fixed-minute boundary you can burst across.  Every response includes:  | Header | Meaning | |---|---| | `X-RateLimit-Limit` | Requests permitted in the current window. | | `X-RateLimit-Remaining` | Requests left in the current window after this call. | | `X-RateLimit-Reset` | Unix epoch (seconds) when the next slot opens. | | `X-RateLimit-Policy` | Machine-readable policy descriptor, e.g. `600;w=60`. | | `Retry-After` | Seconds to wait before retrying. **Only present on 429 responses.** |  **On 429 (rate_limit_exceeded):** the response body matches the standard error envelope with `code: \"rate_limit_exceeded\"`, plus `limit`, `window_seconds`, `retry_after`, and `request_id` fields. SDKs MUST honor `Retry-After` and use exponential backoff with jitter on subsequent retries — never a tight loop.  Recommended backoff: ``` sleep_ms = (Retry-After * 1000) + random(0..250) ```  Monthly + daily-AI tier quotas (`free`, `starter`, `custom`) are enforced separately and also surface as 429s; they include `tier`, `scope`, and `resets_at` fields.  ## Plan Limits (402 — `listings_limit_exceeded`) The Repull API also enforces a per-tier cap on **active listings**:  | Tier | Active listings cap | |---|---| | `free` | 5 | | `starter` | 50 | | `custom` | unlimited |  When a customer's active-listing count is above their tier cap, the API returns **`402 Payment Required`** with `error.code = \"listings_limit_exceeded\"` on every route EXCEPT:  - `/v1/health` — uptime probes are never gated. - `/v1/usage/*` — so dashboards can render the over-cap state. - Any `DELETE` — so the customer can trim listings to get back under the cap without paying.  Unlike 429, 402 is NOT a \"wait and retry\" condition — `Retry-After` is not set. The only paths back to 200 are:   1. `DELETE` enough listings to come back under the cap, or   2. Upgrade at `https://repull.dev/dashboard/billing`. The server-side usage cache is 60s, so the first 200 after an upgrade may take up to a minute.  The envelope mirrors `rate_limit_exceeded` for SDK ergonomics: `tier`, `limit`, `active_listings`, `upgrade_url`, plus the standard `code` / `message` / `fix` / `docs_url` / `request_id`.
 
 The version of the OpenAPI document: 1.0.0
 Contact: ivan@vanio.ai

data/lib/repull/models/webhook_event_type.rb

@@ -1,7 +1,7 @@
 =begin
 #Repull API
 
-#The unified API for vacation rental tech. Connect to 50+ PMS platforms and 4 OTA channels through one REST API. Built-in AI operations for guest communication, pricing, and listing optimization.  ## Designed for AI agents Every error response on this API includes machine-parseable fields so an LLM (Claude in MCP, Cursor, Cline, GPT, etc.) can self-recover without escalating to a human: - `error.code` — stable string identifier (e.g. `invalid_params`, `rate_limit_exceeded`) - `error.message` — human-readable cause - `error.fix` — exact recovery steps (e.g. \"Pass `check_in_after` as ISO 8601: `?check_in_after=2026-01-15`\") - `error.docs_url` — link to the canonical write-up at `https://repull.dev/docs/errors/{code}` - `error.request_id` — id to correlate with server-side logs - `error.field` / `error.value_received` / `error.valid_values` / `error.did_you_mean` — when the error is parameter-specific - `error.retry_after` — seconds to wait before retrying (rate-limit + transient upstream)  `Access-Control-Expose-Headers` lists `x-request-id` and the `X-RateLimit-*` family so browsers can read them on cross-origin responses.  ## Quick Start 1. Get an API key at https://repull.dev/dashboard 2. Connect a PMS: `POST /v1/connect/{provider}` 3. List properties: `GET /v1/properties` 4. Get reservations: `GET /v1/reservations`  ## Authentication All requests require a Bearer token: ``` Authorization: Bearer sk_test_YOUR_API_KEY ```  Sandbox keys start with `sk_test_`, production with `sk_live_`.  ## Request Correlation (X-Request-ID) Every response carries an `X-Request-ID` header, e.g. `X-Request-ID: req_01HXY...`. Include this id in support tickets and bug reports — we can trace the full request lifecycle (auth, rate limit, handler, downstream calls, log row) from a single id.  You may set the header on the inbound request to forward your own trace id; we will echo it back instead of generating a new one. Accepted format: `^[\\\\w.-]{1,128}$`.  The id is also embedded in error envelopes as `request_id` so server-side log diffs work even when the response headers are stripped by an intermediate proxy.  ## Rate Limits The public API enforces a per-API-key sliding-window rate limit on top of the per-tier monthly + daily-AI quotas.  **Default policy:** 600 requests per 60 seconds, per API key. Sliding window — there is no fixed-minute boundary you can burst across.  Every response includes:  | Header | Meaning | |---|---| | `X-RateLimit-Limit` | Requests permitted in the current window. | | `X-RateLimit-Remaining` | Requests left in the current window after this call. | | `X-RateLimit-Reset` | Unix epoch (seconds) when the next slot opens. | | `X-RateLimit-Policy` | Machine-readable policy descriptor, e.g. `600;w=60`. | | `Retry-After` | Seconds to wait before retrying. **Only present on 429 responses.** |  **On 429 (rate_limit_exceeded):** the response body matches the standard error envelope with `code: \"rate_limit_exceeded\"`, plus `limit`, `window_seconds`, `retry_after`, and `request_id` fields. SDKs MUST honor `Retry-After` and use exponential backoff with jitter on subsequent retries — never a tight loop.  Recommended backoff: ``` sleep_ms = (Retry-After * 1000) + random(0..250) ```  Monthly + daily-AI tier quotas (`free`, `starter`, `custom`) are enforced separately and also surface as 429s; they include `tier`, `scope`, and `resets_at` fields.
+#The unified API for vacation rental tech. Connect to 50+ PMS platforms and 4 OTA channels through one REST API. Built-in AI operations for guest communication, pricing, and listing optimization.  ## Designed for AI agents Every error response on this API includes machine-parseable fields so an LLM (Claude in MCP, Cursor, Cline, GPT, etc.) can self-recover without escalating to a human: - `error.code` — stable string identifier (e.g. `invalid_params`, `rate_limit_exceeded`) - `error.message` — human-readable cause - `error.fix` — exact recovery steps (e.g. \"Pass `check_in_after` as ISO 8601: `?check_in_after=2026-01-15`\") - `error.docs_url` — link to the canonical write-up at `https://repull.dev/docs/errors/{code}` - `error.request_id` — id to correlate with server-side logs - `error.field` / `error.value_received` / `error.valid_values` / `error.did_you_mean` — when the error is parameter-specific - `error.retry_after` — seconds to wait before retrying (rate-limit + transient upstream)  `Access-Control-Expose-Headers` lists `x-request-id` and the `X-RateLimit-*` family so browsers can read them on cross-origin responses.  ## Quick Start 1. Get an API key at https://repull.dev/dashboard 2. Connect a PMS: `POST /v1/connect/{provider}` 3. List properties: `GET /v1/properties` 4. Get reservations: `GET /v1/reservations`  ## Authentication All requests require a Bearer token: ``` Authorization: Bearer sk_test_YOUR_API_KEY ```  Sandbox keys start with `sk_test_`, production with `sk_live_`.  ## Request Correlation (X-Request-ID) Every response carries an `X-Request-ID` header, e.g. `X-Request-ID: req_01HXY...`. Include this id in support tickets and bug reports — we can trace the full request lifecycle (auth, rate limit, handler, downstream calls, log row) from a single id.  You may set the header on the inbound request to forward your own trace id; we will echo it back instead of generating a new one. Accepted format: `^[\\\\w.-]{1,128}$`.  The id is also embedded in error envelopes as `request_id` so server-side log diffs work even when the response headers are stripped by an intermediate proxy.  ## Rate Limits The public API enforces a per-API-key sliding-window rate limit on top of the per-tier monthly + daily-AI quotas.  **Default policy:** 600 requests per 60 seconds, per API key. Sliding window — there is no fixed-minute boundary you can burst across.  Every response includes:  | Header | Meaning | |---|---| | `X-RateLimit-Limit` | Requests permitted in the current window. | | `X-RateLimit-Remaining` | Requests left in the current window after this call. | | `X-RateLimit-Reset` | Unix epoch (seconds) when the next slot opens. | | `X-RateLimit-Policy` | Machine-readable policy descriptor, e.g. `600;w=60`. | | `Retry-After` | Seconds to wait before retrying. **Only present on 429 responses.** |  **On 429 (rate_limit_exceeded):** the response body matches the standard error envelope with `code: \"rate_limit_exceeded\"`, plus `limit`, `window_seconds`, `retry_after`, and `request_id` fields. SDKs MUST honor `Retry-After` and use exponential backoff with jitter on subsequent retries — never a tight loop.  Recommended backoff: ``` sleep_ms = (Retry-After * 1000) + random(0..250) ```  Monthly + daily-AI tier quotas (`free`, `starter`, `custom`) are enforced separately and also surface as 429s; they include `tier`, `scope`, and `resets_at` fields.  ## Plan Limits (402 — `listings_limit_exceeded`) The Repull API also enforces a per-tier cap on **active listings**:  | Tier | Active listings cap | |---|---| | `free` | 5 | | `starter` | 50 | | `custom` | unlimited |  When a customer's active-listing count is above their tier cap, the API returns **`402 Payment Required`** with `error.code = \"listings_limit_exceeded\"` on every route EXCEPT:  - `/v1/health` — uptime probes are never gated. - `/v1/usage/*` — so dashboards can render the over-cap state. - Any `DELETE` — so the customer can trim listings to get back under the cap without paying.  Unlike 429, 402 is NOT a \"wait and retry\" condition — `Retry-After` is not set. The only paths back to 200 are:   1. `DELETE` enough listings to come back under the cap, or   2. Upgrade at `https://repull.dev/dashboard/billing`. The server-side usage cache is 60s, so the first 200 after an upgrade may take up to a minute.  The envelope mirrors `rate_limit_exceeded` for SDK ergonomics: `tier`, `limit`, `active_listings`, `upgrade_url`, plus the standard `code` / `message` / `fix` / `docs_url` / `request_id`.
 
 The version of the OpenAPI document: 1.0.0
 Contact: ivan@vanio.ai

data/lib/repull/models/webhook_list_response.rb

@@ -1,7 +1,7 @@
 =begin
 #Repull API
 
-#The unified API for vacation rental tech. Connect to 50+ PMS platforms and 4 OTA channels through one REST API. Built-in AI operations for guest communication, pricing, and listing optimization.  ## Designed for AI agents Every error response on this API includes machine-parseable fields so an LLM (Claude in MCP, Cursor, Cline, GPT, etc.) can self-recover without escalating to a human: - `error.code` — stable string identifier (e.g. `invalid_params`, `rate_limit_exceeded`) - `error.message` — human-readable cause - `error.fix` — exact recovery steps (e.g. \"Pass `check_in_after` as ISO 8601: `?check_in_after=2026-01-15`\") - `error.docs_url` — link to the canonical write-up at `https://repull.dev/docs/errors/{code}` - `error.request_id` — id to correlate with server-side logs - `error.field` / `error.value_received` / `error.valid_values` / `error.did_you_mean` — when the error is parameter-specific - `error.retry_after` — seconds to wait before retrying (rate-limit + transient upstream)  `Access-Control-Expose-Headers` lists `x-request-id` and the `X-RateLimit-*` family so browsers can read them on cross-origin responses.  ## Quick Start 1. Get an API key at https://repull.dev/dashboard 2. Connect a PMS: `POST /v1/connect/{provider}` 3. List properties: `GET /v1/properties` 4. Get reservations: `GET /v1/reservations`  ## Authentication All requests require a Bearer token: ``` Authorization: Bearer sk_test_YOUR_API_KEY ```  Sandbox keys start with `sk_test_`, production with `sk_live_`.  ## Request Correlation (X-Request-ID) Every response carries an `X-Request-ID` header, e.g. `X-Request-ID: req_01HXY...`. Include this id in support tickets and bug reports — we can trace the full request lifecycle (auth, rate limit, handler, downstream calls, log row) from a single id.  You may set the header on the inbound request to forward your own trace id; we will echo it back instead of generating a new one. Accepted format: `^[\\\\w.-]{1,128}$`.  The id is also embedded in error envelopes as `request_id` so server-side log diffs work even when the response headers are stripped by an intermediate proxy.  ## Rate Limits The public API enforces a per-API-key sliding-window rate limit on top of the per-tier monthly + daily-AI quotas.  **Default policy:** 600 requests per 60 seconds, per API key. Sliding window — there is no fixed-minute boundary you can burst across.  Every response includes:  | Header | Meaning | |---|---| | `X-RateLimit-Limit` | Requests permitted in the current window. | | `X-RateLimit-Remaining` | Requests left in the current window after this call. | | `X-RateLimit-Reset` | Unix epoch (seconds) when the next slot opens. | | `X-RateLimit-Policy` | Machine-readable policy descriptor, e.g. `600;w=60`. | | `Retry-After` | Seconds to wait before retrying. **Only present on 429 responses.** |  **On 429 (rate_limit_exceeded):** the response body matches the standard error envelope with `code: \"rate_limit_exceeded\"`, plus `limit`, `window_seconds`, `retry_after`, and `request_id` fields. SDKs MUST honor `Retry-After` and use exponential backoff with jitter on subsequent retries — never a tight loop.  Recommended backoff: ``` sleep_ms = (Retry-After * 1000) + random(0..250) ```  Monthly + daily-AI tier quotas (`free`, `starter`, `custom`) are enforced separately and also surface as 429s; they include `tier`, `scope`, and `resets_at` fields.
+#The unified API for vacation rental tech. Connect to 50+ PMS platforms and 4 OTA channels through one REST API. Built-in AI operations for guest communication, pricing, and listing optimization.  ## Designed for AI agents Every error response on this API includes machine-parseable fields so an LLM (Claude in MCP, Cursor, Cline, GPT, etc.) can self-recover without escalating to a human: - `error.code` — stable string identifier (e.g. `invalid_params`, `rate_limit_exceeded`) - `error.message` — human-readable cause - `error.fix` — exact recovery steps (e.g. \"Pass `check_in_after` as ISO 8601: `?check_in_after=2026-01-15`\") - `error.docs_url` — link to the canonical write-up at `https://repull.dev/docs/errors/{code}` - `error.request_id` — id to correlate with server-side logs - `error.field` / `error.value_received` / `error.valid_values` / `error.did_you_mean` — when the error is parameter-specific - `error.retry_after` — seconds to wait before retrying (rate-limit + transient upstream)  `Access-Control-Expose-Headers` lists `x-request-id` and the `X-RateLimit-*` family so browsers can read them on cross-origin responses.  ## Quick Start 1. Get an API key at https://repull.dev/dashboard 2. Connect a PMS: `POST /v1/connect/{provider}` 3. List properties: `GET /v1/properties` 4. Get reservations: `GET /v1/reservations`  ## Authentication All requests require a Bearer token: ``` Authorization: Bearer sk_test_YOUR_API_KEY ```  Sandbox keys start with `sk_test_`, production with `sk_live_`.  ## Request Correlation (X-Request-ID) Every response carries an `X-Request-ID` header, e.g. `X-Request-ID: req_01HXY...`. Include this id in support tickets and bug reports — we can trace the full request lifecycle (auth, rate limit, handler, downstream calls, log row) from a single id.  You may set the header on the inbound request to forward your own trace id; we will echo it back instead of generating a new one. Accepted format: `^[\\\\w.-]{1,128}$`.  The id is also embedded in error envelopes as `request_id` so server-side log diffs work even when the response headers are stripped by an intermediate proxy.  ## Rate Limits The public API enforces a per-API-key sliding-window rate limit on top of the per-tier monthly + daily-AI quotas.  **Default policy:** 600 requests per 60 seconds, per API key. Sliding window — there is no fixed-minute boundary you can burst across.  Every response includes:  | Header | Meaning | |---|---| | `X-RateLimit-Limit` | Requests permitted in the current window. | | `X-RateLimit-Remaining` | Requests left in the current window after this call. | | `X-RateLimit-Reset` | Unix epoch (seconds) when the next slot opens. | | `X-RateLimit-Policy` | Machine-readable policy descriptor, e.g. `600;w=60`. | | `Retry-After` | Seconds to wait before retrying. **Only present on 429 responses.** |  **On 429 (rate_limit_exceeded):** the response body matches the standard error envelope with `code: \"rate_limit_exceeded\"`, plus `limit`, `window_seconds`, `retry_after`, and `request_id` fields. SDKs MUST honor `Retry-After` and use exponential backoff with jitter on subsequent retries — never a tight loop.  Recommended backoff: ``` sleep_ms = (Retry-After * 1000) + random(0..250) ```  Monthly + daily-AI tier quotas (`free`, `starter`, `custom`) are enforced separately and also surface as 429s; they include `tier`, `scope`, and `resets_at` fields.  ## Plan Limits (402 — `listings_limit_exceeded`) The Repull API also enforces a per-tier cap on **active listings**:  | Tier | Active listings cap | |---|---| | `free` | 5 | | `starter` | 50 | | `custom` | unlimited |  When a customer's active-listing count is above their tier cap, the API returns **`402 Payment Required`** with `error.code = \"listings_limit_exceeded\"` on every route EXCEPT:  - `/v1/health` — uptime probes are never gated. - `/v1/usage/*` — so dashboards can render the over-cap state. - Any `DELETE` — so the customer can trim listings to get back under the cap without paying.  Unlike 429, 402 is NOT a \"wait and retry\" condition — `Retry-After` is not set. The only paths back to 200 are:   1. `DELETE` enough listings to come back under the cap, or   2. Upgrade at `https://repull.dev/dashboard/billing`. The server-side usage cache is 60s, so the first 200 after an upgrade may take up to a minute.  The envelope mirrors `rate_limit_exceeded` for SDK ergonomics: `tier`, `limit`, `active_listings`, `upgrade_url`, plus the standard `code` / `message` / `fix` / `docs_url` / `request_id`.
 
 The version of the OpenAPI document: 1.0.0
 Contact: ivan@vanio.ai

data/lib/repull/models/webhook_subscription.rb

@@ -1,7 +1,7 @@
 =begin
 #Repull API
 
-#The unified API for vacation rental tech. Connect to 50+ PMS platforms and 4 OTA channels through one REST API. Built-in AI operations for guest communication, pricing, and listing optimization.  ## Designed for AI agents Every error response on this API includes machine-parseable fields so an LLM (Claude in MCP, Cursor, Cline, GPT, etc.) can self-recover without escalating to a human: - `error.code` — stable string identifier (e.g. `invalid_params`, `rate_limit_exceeded`) - `error.message` — human-readable cause - `error.fix` — exact recovery steps (e.g. \"Pass `check_in_after` as ISO 8601: `?check_in_after=2026-01-15`\") - `error.docs_url` — link to the canonical write-up at `https://repull.dev/docs/errors/{code}` - `error.request_id` — id to correlate with server-side logs - `error.field` / `error.value_received` / `error.valid_values` / `error.did_you_mean` — when the error is parameter-specific - `error.retry_after` — seconds to wait before retrying (rate-limit + transient upstream)  `Access-Control-Expose-Headers` lists `x-request-id` and the `X-RateLimit-*` family so browsers can read them on cross-origin responses.  ## Quick Start 1. Get an API key at https://repull.dev/dashboard 2. Connect a PMS: `POST /v1/connect/{provider}` 3. List properties: `GET /v1/properties` 4. Get reservations: `GET /v1/reservations`  ## Authentication All requests require a Bearer token: ``` Authorization: Bearer sk_test_YOUR_API_KEY ```  Sandbox keys start with `sk_test_`, production with `sk_live_`.  ## Request Correlation (X-Request-ID) Every response carries an `X-Request-ID` header, e.g. `X-Request-ID: req_01HXY...`. Include this id in support tickets and bug reports — we can trace the full request lifecycle (auth, rate limit, handler, downstream calls, log row) from a single id.  You may set the header on the inbound request to forward your own trace id; we will echo it back instead of generating a new one. Accepted format: `^[\\\\w.-]{1,128}$`.  The id is also embedded in error envelopes as `request_id` so server-side log diffs work even when the response headers are stripped by an intermediate proxy.  ## Rate Limits The public API enforces a per-API-key sliding-window rate limit on top of the per-tier monthly + daily-AI quotas.  **Default policy:** 600 requests per 60 seconds, per API key. Sliding window — there is no fixed-minute boundary you can burst across.  Every response includes:  | Header | Meaning | |---|---| | `X-RateLimit-Limit` | Requests permitted in the current window. | | `X-RateLimit-Remaining` | Requests left in the current window after this call. | | `X-RateLimit-Reset` | Unix epoch (seconds) when the next slot opens. | | `X-RateLimit-Policy` | Machine-readable policy descriptor, e.g. `600;w=60`. | | `Retry-After` | Seconds to wait before retrying. **Only present on 429 responses.** |  **On 429 (rate_limit_exceeded):** the response body matches the standard error envelope with `code: \"rate_limit_exceeded\"`, plus `limit`, `window_seconds`, `retry_after`, and `request_id` fields. SDKs MUST honor `Retry-After` and use exponential backoff with jitter on subsequent retries — never a tight loop.  Recommended backoff: ``` sleep_ms = (Retry-After * 1000) + random(0..250) ```  Monthly + daily-AI tier quotas (`free`, `starter`, `custom`) are enforced separately and also surface as 429s; they include `tier`, `scope`, and `resets_at` fields.
+#The unified API for vacation rental tech. Connect to 50+ PMS platforms and 4 OTA channels through one REST API. Built-in AI operations for guest communication, pricing, and listing optimization.  ## Designed for AI agents Every error response on this API includes machine-parseable fields so an LLM (Claude in MCP, Cursor, Cline, GPT, etc.) can self-recover without escalating to a human: - `error.code` — stable string identifier (e.g. `invalid_params`, `rate_limit_exceeded`) - `error.message` — human-readable cause - `error.fix` — exact recovery steps (e.g. \"Pass `check_in_after` as ISO 8601: `?check_in_after=2026-01-15`\") - `error.docs_url` — link to the canonical write-up at `https://repull.dev/docs/errors/{code}` - `error.request_id` — id to correlate with server-side logs - `error.field` / `error.value_received` / `error.valid_values` / `error.did_you_mean` — when the error is parameter-specific - `error.retry_after` — seconds to wait before retrying (rate-limit + transient upstream)  `Access-Control-Expose-Headers` lists `x-request-id` and the `X-RateLimit-*` family so browsers can read them on cross-origin responses.  ## Quick Start 1. Get an API key at https://repull.dev/dashboard 2. Connect a PMS: `POST /v1/connect/{provider}` 3. List properties: `GET /v1/properties` 4. Get reservations: `GET /v1/reservations`  ## Authentication All requests require a Bearer token: ``` Authorization: Bearer sk_test_YOUR_API_KEY ```  Sandbox keys start with `sk_test_`, production with `sk_live_`.  ## Request Correlation (X-Request-ID) Every response carries an `X-Request-ID` header, e.g. `X-Request-ID: req_01HXY...`. Include this id in support tickets and bug reports — we can trace the full request lifecycle (auth, rate limit, handler, downstream calls, log row) from a single id.  You may set the header on the inbound request to forward your own trace id; we will echo it back instead of generating a new one. Accepted format: `^[\\\\w.-]{1,128}$`.  The id is also embedded in error envelopes as `request_id` so server-side log diffs work even when the response headers are stripped by an intermediate proxy.  ## Rate Limits The public API enforces a per-API-key sliding-window rate limit on top of the per-tier monthly + daily-AI quotas.  **Default policy:** 600 requests per 60 seconds, per API key. Sliding window — there is no fixed-minute boundary you can burst across.  Every response includes:  | Header | Meaning | |---|---| | `X-RateLimit-Limit` | Requests permitted in the current window. | | `X-RateLimit-Remaining` | Requests left in the current window after this call. | | `X-RateLimit-Reset` | Unix epoch (seconds) when the next slot opens. | | `X-RateLimit-Policy` | Machine-readable policy descriptor, e.g. `600;w=60`. | | `Retry-After` | Seconds to wait before retrying. **Only present on 429 responses.** |  **On 429 (rate_limit_exceeded):** the response body matches the standard error envelope with `code: \"rate_limit_exceeded\"`, plus `limit`, `window_seconds`, `retry_after`, and `request_id` fields. SDKs MUST honor `Retry-After` and use exponential backoff with jitter on subsequent retries — never a tight loop.  Recommended backoff: ``` sleep_ms = (Retry-After * 1000) + random(0..250) ```  Monthly + daily-AI tier quotas (`free`, `starter`, `custom`) are enforced separately and also surface as 429s; they include `tier`, `scope`, and `resets_at` fields.  ## Plan Limits (402 — `listings_limit_exceeded`) The Repull API also enforces a per-tier cap on **active listings**:  | Tier | Active listings cap | |---|---| | `free` | 5 | | `starter` | 50 | | `custom` | unlimited |  When a customer's active-listing count is above their tier cap, the API returns **`402 Payment Required`** with `error.code = \"listings_limit_exceeded\"` on every route EXCEPT:  - `/v1/health` — uptime probes are never gated. - `/v1/usage/*` — so dashboards can render the over-cap state. - Any `DELETE` — so the customer can trim listings to get back under the cap without paying.  Unlike 429, 402 is NOT a \"wait and retry\" condition — `Retry-After` is not set. The only paths back to 200 are:   1. `DELETE` enough listings to come back under the cap, or   2. Upgrade at `https://repull.dev/dashboard/billing`. The server-side usage cache is 60s, so the first 200 after an upgrade may take up to a minute.  The envelope mirrors `rate_limit_exceeded` for SDK ergonomics: `tier`, `limit`, `active_listings`, `upgrade_url`, plus the standard `code` / `message` / `fix` / `docs_url` / `request_id`.
 
 The version of the OpenAPI document: 1.0.0
 Contact: ivan@vanio.ai

data/lib/repull/version.rb

@@ -1,7 +1,7 @@
 =begin
 #Repull API
 
-#The unified API for vacation rental tech. Connect to 50+ PMS platforms and 4 OTA channels through one REST API. Built-in AI operations for guest communication, pricing, and listing optimization.  ## Designed for AI agents Every error response on this API includes machine-parseable fields so an LLM (Claude in MCP, Cursor, Cline, GPT, etc.) can self-recover without escalating to a human: - `error.code` — stable string identifier (e.g. `invalid_params`, `rate_limit_exceeded`) - `error.message` — human-readable cause - `error.fix` — exact recovery steps (e.g. \"Pass `check_in_after` as ISO 8601: `?check_in_after=2026-01-15`\") - `error.docs_url` — link to the canonical write-up at `https://repull.dev/docs/errors/{code}` - `error.request_id` — id to correlate with server-side logs - `error.field` / `error.value_received` / `error.valid_values` / `error.did_you_mean` — when the error is parameter-specific - `error.retry_after` — seconds to wait before retrying (rate-limit + transient upstream)  `Access-Control-Expose-Headers` lists `x-request-id` and the `X-RateLimit-*` family so browsers can read them on cross-origin responses.  ## Quick Start 1. Get an API key at https://repull.dev/dashboard 2. Connect a PMS: `POST /v1/connect/{provider}` 3. List properties: `GET /v1/properties` 4. Get reservations: `GET /v1/reservations`  ## Authentication All requests require a Bearer token: ``` Authorization: Bearer sk_test_YOUR_API_KEY ```  Sandbox keys start with `sk_test_`, production with `sk_live_`.  ## Request Correlation (X-Request-ID) Every response carries an `X-Request-ID` header, e.g. `X-Request-ID: req_01HXY...`. Include this id in support tickets and bug reports — we can trace the full request lifecycle (auth, rate limit, handler, downstream calls, log row) from a single id.  You may set the header on the inbound request to forward your own trace id; we will echo it back instead of generating a new one. Accepted format: `^[\\\\w.-]{1,128}$`.  The id is also embedded in error envelopes as `request_id` so server-side log diffs work even when the response headers are stripped by an intermediate proxy.  ## Rate Limits The public API enforces a per-API-key sliding-window rate limit on top of the per-tier monthly + daily-AI quotas.  **Default policy:** 600 requests per 60 seconds, per API key. Sliding window — there is no fixed-minute boundary you can burst across.  Every response includes:  | Header | Meaning | |---|---| | `X-RateLimit-Limit` | Requests permitted in the current window. | | `X-RateLimit-Remaining` | Requests left in the current window after this call. | | `X-RateLimit-Reset` | Unix epoch (seconds) when the next slot opens. | | `X-RateLimit-Policy` | Machine-readable policy descriptor, e.g. `600;w=60`. | | `Retry-After` | Seconds to wait before retrying. **Only present on 429 responses.** |  **On 429 (rate_limit_exceeded):** the response body matches the standard error envelope with `code: \"rate_limit_exceeded\"`, plus `limit`, `window_seconds`, `retry_after`, and `request_id` fields. SDKs MUST honor `Retry-After` and use exponential backoff with jitter on subsequent retries — never a tight loop.  Recommended backoff: ``` sleep_ms = (Retry-After * 1000) + random(0..250) ```  Monthly + daily-AI tier quotas (`free`, `starter`, `custom`) are enforced separately and also surface as 429s; they include `tier`, `scope`, and `resets_at` fields.
+#The unified API for vacation rental tech. Connect to 50+ PMS platforms and 4 OTA channels through one REST API. Built-in AI operations for guest communication, pricing, and listing optimization.  ## Designed for AI agents Every error response on this API includes machine-parseable fields so an LLM (Claude in MCP, Cursor, Cline, GPT, etc.) can self-recover without escalating to a human: - `error.code` — stable string identifier (e.g. `invalid_params`, `rate_limit_exceeded`) - `error.message` — human-readable cause - `error.fix` — exact recovery steps (e.g. \"Pass `check_in_after` as ISO 8601: `?check_in_after=2026-01-15`\") - `error.docs_url` — link to the canonical write-up at `https://repull.dev/docs/errors/{code}` - `error.request_id` — id to correlate with server-side logs - `error.field` / `error.value_received` / `error.valid_values` / `error.did_you_mean` — when the error is parameter-specific - `error.retry_after` — seconds to wait before retrying (rate-limit + transient upstream)  `Access-Control-Expose-Headers` lists `x-request-id` and the `X-RateLimit-*` family so browsers can read them on cross-origin responses.  ## Quick Start 1. Get an API key at https://repull.dev/dashboard 2. Connect a PMS: `POST /v1/connect/{provider}` 3. List properties: `GET /v1/properties` 4. Get reservations: `GET /v1/reservations`  ## Authentication All requests require a Bearer token: ``` Authorization: Bearer sk_test_YOUR_API_KEY ```  Sandbox keys start with `sk_test_`, production with `sk_live_`.  ## Request Correlation (X-Request-ID) Every response carries an `X-Request-ID` header, e.g. `X-Request-ID: req_01HXY...`. Include this id in support tickets and bug reports — we can trace the full request lifecycle (auth, rate limit, handler, downstream calls, log row) from a single id.  You may set the header on the inbound request to forward your own trace id; we will echo it back instead of generating a new one. Accepted format: `^[\\\\w.-]{1,128}$`.  The id is also embedded in error envelopes as `request_id` so server-side log diffs work even when the response headers are stripped by an intermediate proxy.  ## Rate Limits The public API enforces a per-API-key sliding-window rate limit on top of the per-tier monthly + daily-AI quotas.  **Default policy:** 600 requests per 60 seconds, per API key. Sliding window — there is no fixed-minute boundary you can burst across.  Every response includes:  | Header | Meaning | |---|---| | `X-RateLimit-Limit` | Requests permitted in the current window. | | `X-RateLimit-Remaining` | Requests left in the current window after this call. | | `X-RateLimit-Reset` | Unix epoch (seconds) when the next slot opens. | | `X-RateLimit-Policy` | Machine-readable policy descriptor, e.g. `600;w=60`. | | `Retry-After` | Seconds to wait before retrying. **Only present on 429 responses.** |  **On 429 (rate_limit_exceeded):** the response body matches the standard error envelope with `code: \"rate_limit_exceeded\"`, plus `limit`, `window_seconds`, `retry_after`, and `request_id` fields. SDKs MUST honor `Retry-After` and use exponential backoff with jitter on subsequent retries — never a tight loop.  Recommended backoff: ``` sleep_ms = (Retry-After * 1000) + random(0..250) ```  Monthly + daily-AI tier quotas (`free`, `starter`, `custom`) are enforced separately and also surface as 429s; they include `tier`, `scope`, and `resets_at` fields.  ## Plan Limits (402 — `listings_limit_exceeded`) The Repull API also enforces a per-tier cap on **active listings**:  | Tier | Active listings cap | |---|---| | `free` | 5 | | `starter` | 50 | | `custom` | unlimited |  When a customer's active-listing count is above their tier cap, the API returns **`402 Payment Required`** with `error.code = \"listings_limit_exceeded\"` on every route EXCEPT:  - `/v1/health` — uptime probes are never gated. - `/v1/usage/*` — so dashboards can render the over-cap state. - Any `DELETE` — so the customer can trim listings to get back under the cap without paying.  Unlike 429, 402 is NOT a \"wait and retry\" condition — `Retry-After` is not set. The only paths back to 200 are:   1. `DELETE` enough listings to come back under the cap, or   2. Upgrade at `https://repull.dev/dashboard/billing`. The server-side usage cache is 60s, so the first 200 after an upgrade may take up to a minute.  The envelope mirrors `rate_limit_exceeded` for SDK ergonomics: `tier`, `limit`, `active_listings`, `upgrade_url`, plus the standard `code` / `message` / `fix` / `docs_url` / `request_id`.
 
 The version of the OpenAPI document: 1.0.0
 Contact: ivan@vanio.ai
@@ -11,5 +11,5 @@ Generator version: 7.22.0
 =end
 
 module Repull
-  VERSION = '0.2.3'
+  VERSION = '0.2.4'
 end

data/lib/repull.rb

@@ -1,7 +1,7 @@
 =begin
 #Repull API
 
-#The unified API for vacation rental tech. Connect to 50+ PMS platforms and 4 OTA channels through one REST API. Built-in AI operations for guest communication, pricing, and listing optimization.  ## Designed for AI agents Every error response on this API includes machine-parseable fields so an LLM (Claude in MCP, Cursor, Cline, GPT, etc.) can self-recover without escalating to a human: - `error.code` — stable string identifier (e.g. `invalid_params`, `rate_limit_exceeded`) - `error.message` — human-readable cause - `error.fix` — exact recovery steps (e.g. \"Pass `check_in_after` as ISO 8601: `?check_in_after=2026-01-15`\") - `error.docs_url` — link to the canonical write-up at `https://repull.dev/docs/errors/{code}` - `error.request_id` — id to correlate with server-side logs - `error.field` / `error.value_received` / `error.valid_values` / `error.did_you_mean` — when the error is parameter-specific - `error.retry_after` — seconds to wait before retrying (rate-limit + transient upstream)  `Access-Control-Expose-Headers` lists `x-request-id` and the `X-RateLimit-*` family so browsers can read them on cross-origin responses.  ## Quick Start 1. Get an API key at https://repull.dev/dashboard 2. Connect a PMS: `POST /v1/connect/{provider}` 3. List properties: `GET /v1/properties` 4. Get reservations: `GET /v1/reservations`  ## Authentication All requests require a Bearer token: ``` Authorization: Bearer sk_test_YOUR_API_KEY ```  Sandbox keys start with `sk_test_`, production with `sk_live_`.  ## Request Correlation (X-Request-ID) Every response carries an `X-Request-ID` header, e.g. `X-Request-ID: req_01HXY...`. Include this id in support tickets and bug reports — we can trace the full request lifecycle (auth, rate limit, handler, downstream calls, log row) from a single id.  You may set the header on the inbound request to forward your own trace id; we will echo it back instead of generating a new one. Accepted format: `^[\\\\w.-]{1,128}$`.  The id is also embedded in error envelopes as `request_id` so server-side log diffs work even when the response headers are stripped by an intermediate proxy.  ## Rate Limits The public API enforces a per-API-key sliding-window rate limit on top of the per-tier monthly + daily-AI quotas.  **Default policy:** 600 requests per 60 seconds, per API key. Sliding window — there is no fixed-minute boundary you can burst across.  Every response includes:  | Header | Meaning | |---|---| | `X-RateLimit-Limit` | Requests permitted in the current window. | | `X-RateLimit-Remaining` | Requests left in the current window after this call. | | `X-RateLimit-Reset` | Unix epoch (seconds) when the next slot opens. | | `X-RateLimit-Policy` | Machine-readable policy descriptor, e.g. `600;w=60`. | | `Retry-After` | Seconds to wait before retrying. **Only present on 429 responses.** |  **On 429 (rate_limit_exceeded):** the response body matches the standard error envelope with `code: \"rate_limit_exceeded\"`, plus `limit`, `window_seconds`, `retry_after`, and `request_id` fields. SDKs MUST honor `Retry-After` and use exponential backoff with jitter on subsequent retries — never a tight loop.  Recommended backoff: ``` sleep_ms = (Retry-After * 1000) + random(0..250) ```  Monthly + daily-AI tier quotas (`free`, `starter`, `custom`) are enforced separately and also surface as 429s; they include `tier`, `scope`, and `resets_at` fields.
+#The unified API for vacation rental tech. Connect to 50+ PMS platforms and 4 OTA channels through one REST API. Built-in AI operations for guest communication, pricing, and listing optimization.  ## Designed for AI agents Every error response on this API includes machine-parseable fields so an LLM (Claude in MCP, Cursor, Cline, GPT, etc.) can self-recover without escalating to a human: - `error.code` — stable string identifier (e.g. `invalid_params`, `rate_limit_exceeded`) - `error.message` — human-readable cause - `error.fix` — exact recovery steps (e.g. \"Pass `check_in_after` as ISO 8601: `?check_in_after=2026-01-15`\") - `error.docs_url` — link to the canonical write-up at `https://repull.dev/docs/errors/{code}` - `error.request_id` — id to correlate with server-side logs - `error.field` / `error.value_received` / `error.valid_values` / `error.did_you_mean` — when the error is parameter-specific - `error.retry_after` — seconds to wait before retrying (rate-limit + transient upstream)  `Access-Control-Expose-Headers` lists `x-request-id` and the `X-RateLimit-*` family so browsers can read them on cross-origin responses.  ## Quick Start 1. Get an API key at https://repull.dev/dashboard 2. Connect a PMS: `POST /v1/connect/{provider}` 3. List properties: `GET /v1/properties` 4. Get reservations: `GET /v1/reservations`  ## Authentication All requests require a Bearer token: ``` Authorization: Bearer sk_test_YOUR_API_KEY ```  Sandbox keys start with `sk_test_`, production with `sk_live_`.  ## Request Correlation (X-Request-ID) Every response carries an `X-Request-ID` header, e.g. `X-Request-ID: req_01HXY...`. Include this id in support tickets and bug reports — we can trace the full request lifecycle (auth, rate limit, handler, downstream calls, log row) from a single id.  You may set the header on the inbound request to forward your own trace id; we will echo it back instead of generating a new one. Accepted format: `^[\\\\w.-]{1,128}$`.  The id is also embedded in error envelopes as `request_id` so server-side log diffs work even when the response headers are stripped by an intermediate proxy.  ## Rate Limits The public API enforces a per-API-key sliding-window rate limit on top of the per-tier monthly + daily-AI quotas.  **Default policy:** 600 requests per 60 seconds, per API key. Sliding window — there is no fixed-minute boundary you can burst across.  Every response includes:  | Header | Meaning | |---|---| | `X-RateLimit-Limit` | Requests permitted in the current window. | | `X-RateLimit-Remaining` | Requests left in the current window after this call. | | `X-RateLimit-Reset` | Unix epoch (seconds) when the next slot opens. | | `X-RateLimit-Policy` | Machine-readable policy descriptor, e.g. `600;w=60`. | | `Retry-After` | Seconds to wait before retrying. **Only present on 429 responses.** |  **On 429 (rate_limit_exceeded):** the response body matches the standard error envelope with `code: \"rate_limit_exceeded\"`, plus `limit`, `window_seconds`, `retry_after`, and `request_id` fields. SDKs MUST honor `Retry-After` and use exponential backoff with jitter on subsequent retries — never a tight loop.  Recommended backoff: ``` sleep_ms = (Retry-After * 1000) + random(0..250) ```  Monthly + daily-AI tier quotas (`free`, `starter`, `custom`) are enforced separately and also surface as 429s; they include `tier`, `scope`, and `resets_at` fields.  ## Plan Limits (402 — `listings_limit_exceeded`) The Repull API also enforces a per-tier cap on **active listings**:  | Tier | Active listings cap | |---|---| | `free` | 5 | | `starter` | 50 | | `custom` | unlimited |  When a customer's active-listing count is above their tier cap, the API returns **`402 Payment Required`** with `error.code = \"listings_limit_exceeded\"` on every route EXCEPT:  - `/v1/health` — uptime probes are never gated. - `/v1/usage/*` — so dashboards can render the over-cap state. - Any `DELETE` — so the customer can trim listings to get back under the cap without paying.  Unlike 429, 402 is NOT a \"wait and retry\" condition — `Retry-After` is not set. The only paths back to 200 are:   1. `DELETE` enough listings to come back under the cap, or   2. Upgrade at `https://repull.dev/dashboard/billing`. The server-side usage cache is 60s, so the first 200 after an upgrade may take up to a minute.  The envelope mirrors `rate_limit_exceeded` for SDK ergonomics: `tier`, `limit`, `active_listings`, `upgrade_url`, plus the standard `code` / `message` / `fix` / `docs_url` / `request_id`.
 
 The version of the OpenAPI document: 1.0.0
 Contact: ivan@vanio.ai

data/openapi/v1.json

@@ -3,7 +3,7 @@
   "info": {
     "title": "Repull API",
     "version": "1.0.0",
-    "description": "The unified API for vacation rental tech. Connect to 50+ PMS platforms and 4 OTA channels through one REST API. Built-in AI operations for guest communication, pricing, and listing optimization.\n\n## Designed for AI agents\nEvery error response on this API includes machine-parseable fields so an LLM (Claude in MCP, Cursor, Cline, GPT, etc.) can self-recover without escalating to a human:\n- `error.code` \u2014 stable string identifier (e.g. `invalid_params`, `rate_limit_exceeded`)\n- `error.message` \u2014 human-readable cause\n- `error.fix` \u2014 exact recovery steps (e.g. \"Pass `check_in_after` as ISO 8601: `?check_in_after=2026-01-15`\")\n- `error.docs_url` \u2014 link to the canonical write-up at `https://repull.dev/docs/errors/{code}`\n- `error.request_id` \u2014 id to correlate with server-side logs\n- `error.field` / `error.value_received` / `error.valid_values` / `error.did_you_mean` \u2014 when the error is parameter-specific\n- `error.retry_after` \u2014 seconds to wait before retrying (rate-limit + transient upstream)\n\n`Access-Control-Expose-Headers` lists `x-request-id` and the `X-RateLimit-*` family so browsers can read them on cross-origin responses.\n\n## Quick Start\n1. Get an API key at https://repull.dev/dashboard\n2. Connect a PMS: `POST /v1/connect/{provider}`\n3. List properties: `GET /v1/properties`\n4. Get reservations: `GET /v1/reservations`\n\n## Authentication\nAll requests require a Bearer token:\n```\nAuthorization: Bearer sk_test_YOUR_API_KEY\n```\n\nSandbox keys start with `sk_test_`, production with `sk_live_`.\n\n## Request Correlation (X-Request-ID)\nEvery response carries an `X-Request-ID` header, e.g. `X-Request-ID: req_01HXY...`. Include this id in support tickets and bug reports \u2014 we can trace the full request lifecycle (auth, rate limit, handler, downstream calls, log row) from a single id.\n\nYou may set the header on the inbound request to forward your own trace id; we will echo it back instead of generating a new one. Accepted format: `^[\\\\w.-]{1,128}$`.\n\nThe id is also embedded in error envelopes as `request_id` so server-side log diffs work even when the response headers are stripped by an intermediate proxy.\n\n## Rate Limits\nThe public API enforces a per-API-key sliding-window rate limit on top of the per-tier monthly + daily-AI quotas.\n\n**Default policy:** 600 requests per 60 seconds, per API key. Sliding window \u2014 there is no fixed-minute boundary you can burst across.\n\nEvery response includes:\n\n| Header | Meaning |\n|---|---|\n| `X-RateLimit-Limit` | Requests permitted in the current window. |\n| `X-RateLimit-Remaining` | Requests left in the current window after this call. |\n| `X-RateLimit-Reset` | Unix epoch (seconds) when the next slot opens. |\n| `X-RateLimit-Policy` | Machine-readable policy descriptor, e.g. `600;w=60`. |\n| `Retry-After` | Seconds to wait before retrying. **Only present on 429 responses.** |\n\n**On 429 (rate_limit_exceeded):** the response body matches the standard error envelope with `code: \"rate_limit_exceeded\"`, plus `limit`, `window_seconds`, `retry_after`, and `request_id` fields. SDKs MUST honor `Retry-After` and use exponential backoff with jitter on subsequent retries \u2014 never a tight loop.\n\nRecommended backoff:\n```\nsleep_ms = (Retry-After * 1000) + random(0..250)\n```\n\nMonthly + daily-AI tier quotas (`free`, `starter`, `custom`) are enforced separately and also surface as 429s; they include `tier`, `scope`, and `resets_at` fields.",
+    "description": "The unified API for vacation rental tech. Connect to 50+ PMS platforms and 4 OTA channels through one REST API. Built-in AI operations for guest communication, pricing, and listing optimization.\n\n## Designed for AI agents\nEvery error response on this API includes machine-parseable fields so an LLM (Claude in MCP, Cursor, Cline, GPT, etc.) can self-recover without escalating to a human:\n- `error.code` \u2014 stable string identifier (e.g. `invalid_params`, `rate_limit_exceeded`)\n- `error.message` \u2014 human-readable cause\n- `error.fix` \u2014 exact recovery steps (e.g. \"Pass `check_in_after` as ISO 8601: `?check_in_after=2026-01-15`\")\n- `error.docs_url` \u2014 link to the canonical write-up at `https://repull.dev/docs/errors/{code}`\n- `error.request_id` \u2014 id to correlate with server-side logs\n- `error.field` / `error.value_received` / `error.valid_values` / `error.did_you_mean` \u2014 when the error is parameter-specific\n- `error.retry_after` \u2014 seconds to wait before retrying (rate-limit + transient upstream)\n\n`Access-Control-Expose-Headers` lists `x-request-id` and the `X-RateLimit-*` family so browsers can read them on cross-origin responses.\n\n## Quick Start\n1. Get an API key at https://repull.dev/dashboard\n2. Connect a PMS: `POST /v1/connect/{provider}`\n3. List properties: `GET /v1/properties`\n4. Get reservations: `GET /v1/reservations`\n\n## Authentication\nAll requests require a Bearer token:\n```\nAuthorization: Bearer sk_test_YOUR_API_KEY\n```\n\nSandbox keys start with `sk_test_`, production with `sk_live_`.\n\n## Request Correlation (X-Request-ID)\nEvery response carries an `X-Request-ID` header, e.g. `X-Request-ID: req_01HXY...`. Include this id in support tickets and bug reports \u2014 we can trace the full request lifecycle (auth, rate limit, handler, downstream calls, log row) from a single id.\n\nYou may set the header on the inbound request to forward your own trace id; we will echo it back instead of generating a new one. Accepted format: `^[\\\\w.-]{1,128}$`.\n\nThe id is also embedded in error envelopes as `request_id` so server-side log diffs work even when the response headers are stripped by an intermediate proxy.\n\n## Rate Limits\nThe public API enforces a per-API-key sliding-window rate limit on top of the per-tier monthly + daily-AI quotas.\n\n**Default policy:** 600 requests per 60 seconds, per API key. Sliding window \u2014 there is no fixed-minute boundary you can burst across.\n\nEvery response includes:\n\n| Header | Meaning |\n|---|---|\n| `X-RateLimit-Limit` | Requests permitted in the current window. |\n| `X-RateLimit-Remaining` | Requests left in the current window after this call. |\n| `X-RateLimit-Reset` | Unix epoch (seconds) when the next slot opens. |\n| `X-RateLimit-Policy` | Machine-readable policy descriptor, e.g. `600;w=60`. |\n| `Retry-After` | Seconds to wait before retrying. **Only present on 429 responses.** |\n\n**On 429 (rate_limit_exceeded):** the response body matches the standard error envelope with `code: \"rate_limit_exceeded\"`, plus `limit`, `window_seconds`, `retry_after`, and `request_id` fields. SDKs MUST honor `Retry-After` and use exponential backoff with jitter on subsequent retries \u2014 never a tight loop.\n\nRecommended backoff:\n```\nsleep_ms = (Retry-After * 1000) + random(0..250)\n```\n\nMonthly + daily-AI tier quotas (`free`, `starter`, `custom`) are enforced separately and also surface as 429s; they include `tier`, `scope`, and `resets_at` fields.\n\n## Plan Limits (402 \u2014 `listings_limit_exceeded`)\nThe Repull API also enforces a per-tier cap on **active listings**:\n\n| Tier | Active listings cap |\n|---|---|\n| `free` | 5 |\n| `starter` | 50 |\n| `custom` | unlimited |\n\nWhen a customer's active-listing count is above their tier cap, the API returns **`402 Payment Required`** with `error.code = \"listings_limit_exceeded\"` on every route EXCEPT:\n\n- `/v1/health` \u2014 uptime probes are never gated.\n- `/v1/usage/*` \u2014 so dashboards can render the over-cap state.\n- Any `DELETE` \u2014 so the customer can trim listings to get back under the cap without paying.\n\nUnlike 429, 402 is NOT a \"wait and retry\" condition \u2014 `Retry-After` is not set. The only paths back to 200 are:\n  1. `DELETE` enough listings to come back under the cap, or\n  2. Upgrade at `https://repull.dev/dashboard/billing`. The server-side usage cache is 60s, so the first 200 after an upgrade may take up to a minute.\n\nThe envelope mirrors `rate_limit_exceeded` for SDK ergonomics: `tier`, `limit`, `active_listings`, `upgrade_url`, plus the standard `code` / `message` / `fix` / `docs_url` / `request_id`.",
     "contact": {
       "email": "ivan@vanio.ai",
       "url": "https://repull.dev"
@@ -6274,6 +6274,34 @@
           }
         }
       },
+      "PaymentRequired": {
+        "description": "Plan-listings cap exceeded. The customer's active listing count is above the cap of their resolved Repull tier (`free` = 5, `starter` = 50, `custom` = unlimited). Returned on every route except `/v1/health`, `/v1/usage/*`, and any `DELETE` \u2014 these stay served so the customer can read their state, render the over-cap UI, or trim listings to get back under the cap.\n\nDistinct from 429 / `rate_limit_exceeded` \u2014 this is NOT a \"wait and retry\" condition. The only paths back to a 200 are:\n  1. `DELETE` enough listings via `DELETE /v1/listings/{id}` (or the channel-specific listings DELETE) until `active_listings <= limit`.\n  2. Upgrade to a tier with a higher cap at `https://repull.dev/dashboard/billing`. The very next request after the upgrade lands will see the new cap (server-side usage cache is 60s, so allow up to 1 minute for propagation).\n\nNo `Retry-After` header \u2014 backoff doesn't fix this.",
+        "content": {
+          "application/json": {
+            "schema": {
+              "$ref": "#/components/schemas/Error"
+            },
+            "examples": {
+              "freeTierOverCap": {
+                "summary": "Free tier customer over the 5-listings cap",
+                "value": {
+                  "error": {
+                    "code": "listings_limit_exceeded",
+                    "message": "Your account has 24 active listings but the 'free' tier is capped at 5. The Repull API only serves accounts whose listing count is within their plan.",
+                    "fix": "Either reduce your active listings to 5 or fewer (via DELETE endpoints \u2014 these are still served), or upgrade your plan at https://repull.dev/dashboard/billing to lift the cap immediately.",
+                    "docs_url": "https://repull.dev/docs/errors/listings_limit_exceeded",
+                    "request_id": "req_01J5X7Y8Z9ABCDEF12345678",
+                    "tier": "free",
+                    "limit": 5,
+                    "active_listings": 24,
+                    "upgrade_url": "https://repull.dev/dashboard/billing"
+                  }
+                }
+              }
+            }
+          }
+        }
+      },
       "TooManyRequests": {
         "description": "Quota exceeded. Wait `retry_after` seconds (or until `X-RateLimit-Reset`) before retrying.",
         "headers": {

metadata

@@ -1,14 +1,14 @@
 --- !ruby/object:Gem::Specification
 name: repull
 version: !ruby/object:Gem::Version
-  version: 0.2.3
+  version: 0.2.4
 platform: ruby
 authors:
 - Ivan Nikolov
 autorequire:
 bindir: bin
 cert_chain: []
-date: 2026-05-10 00:00:00.000000000 Z
+date: 2026-05-15 00:00:00.000000000 Z
 dependencies:
 - !ruby/object:Gem::Dependency
   name: typhoeus