{"id":"org_XCtsUSrelarAybkAnBD3g","slug":"openai","name":"OpenAI","domain":"openai.com","description":null,"category":"ai","tags":["api","gpt","llm"],"sourceCount":7,"releaseCount":1008,"releasesLast30Days":67,"avgReleasesPerWeek":18.2,"lastFetchedAt":"2026-04-19T03:01:59.411Z","trackingSince":"2024-05-29T00:00:00.000Z","aliases":[],"accounts":[{"platform":"github","handle":"openai"}],"products":[{"id":"prod_620id5eCzPlmwV1xRa-Q9","slug":"api","name":"API","url":null,"description":"OpenAI platform API and developer tools","sourceCount":1},{"id":"prod_oXXFL920GjpCB1qBWELIT","slug":"chatgpt","name":"ChatGPT","url":null,"description":null,"sourceCount":0},{"id":"prod_0iP-smZKtQJYKB6w0zrye","slug":"codex","name":"Codex","url":null,"description":null,"sourceCount":1},{"id":"prod_mOmj5srOVYLLRS2b0ZPR7","slug":"gpt-api","name":"GPT API","url":null,"description":null,"sourceCount":0}],"sources":[{"id":"src_8tt50oaCp5HlX5LowU2Cb","slug":"openai-net-sdk","name":".NET SDK","type":"github","url":"https://github.com/openai/openai-dotnet","isPrimary":false,"isHidden":false,"fetchPriority":"low","releaseCount":31,"latestVersion":"OpenAI_2.10.0","latestDate":"2026-04-04T00:04:49.000Z","latestAddedAt":"2026-04-07T17:28:00.959Z","productSlug":null,"productName":null},{"id":"src_lUUecdtTa2GdtjHTQGpjm","slug":"openai-api-changelog","name":"API Changelog","type":"feed","url":"https://developers.openai.com/api/docs/changelog","isPrimary":false,"isHidden":false,"fetchPriority":"normal","releaseCount":128,"latestVersion":null,"latestDate":"2026-03-12T00:00:00.000Z","latestAddedAt":"2026-04-07T23:01:40.072Z","productSlug":"api","productName":"API"},{"id":"src_5o4Llv42g_j-oLZvOjA9p","slug":"openai-codex-changelog","name":"Codex Changelog","type":"feed","url":"https://developers.openai.com/codex/changelog","isPrimary":false,"isHidden":false,"fetchPriority":"normal","releaseCount":59,"latestVersion":null,"latestDate":"2026-04-16T00:00:00.000Z","latestAddedAt":"2026-04-16T20:05:11.184Z","productSlug":"codex","productName":"Codex"},{"id":"src_Mqu8-CHN7z7Mii2OjcKKX","slug":"openai-go-sdk","name":"Go SDK","type":"github","url":"https://github.com/openai/openai-go","isPrimary":false,"isHidden":false,"fetchPriority":"normal","releaseCount":154,"latestVersion":"v3.32.0","latestDate":"2026-04-16T17:42:47.000Z","latestAddedAt":"2026-04-16T20:04:08.529Z","productSlug":null,"productName":null},{"id":"src_E8SOv7oaCXxshXd3BtEii","slug":"openai-node-sdk","name":"Node SDK","type":"github","url":"https://github.com/openai/openai-node","isPrimary":false,"isHidden":false,"fetchPriority":"normal","releaseCount":200,"latestVersion":"v6.34.0","latestDate":"2026-04-08T21:26:04.000Z","latestAddedAt":"2026-04-15T01:12:57.721Z","productSlug":null,"productName":null},{"id":"src_9GF-s3jI4ndAEnclz5Yku","slug":"openai-news","name":"Product Releases","type":"feed","url":"https://openai.com/news/product-releases/","isPrimary":true,"isHidden":false,"fetchPriority":"normal","releaseCount":235,"latestVersion":null,"latestDate":"2026-04-16T10:00:00.000Z","latestAddedAt":"2026-04-16T20:04:40.876Z","productSlug":null,"productName":null},{"id":"src_jsbnrYBzcKUcWTVRDsop2","slug":"openai-python-sdk","name":"Python SDK","type":"github","url":"https://github.com/openai/openai-python","isPrimary":false,"isHidden":false,"fetchPriority":"normal","releaseCount":201,"latestVersion":"v2.32.0","latestDate":"2026-04-15T22:27:40.000Z","latestAddedAt":"2026-04-16T02:02:20.852Z","productSlug":null,"productName":null}],"overview":{"scope":"org","content":"**Agents SDK shipped native sandbox execution and expanded Codex with in-app browser, computer use, and marketplace plugins.**\n\n**Go, Node, and Python SDKs aligned on message execution tracking and ephemeral token security.** Go (`v3.32.0`), Node (`v6.34.0`), and Python (`v2.32.0`) added `phase` field to conversation messages for pipeline visibility into execution stages. All three implemented **short-lived token support** to reduce credential exposure in temporary sessions. Go and Node formalized `OAuthErrorCode` types; Go and Python added `detail` to `InputFileContent`. Node (`v6.33.0`-`v6.34.0`) extended async iterators and stream support to WebSocket classes. Python added WebSocket reconnection handlers and raw data transmission.\n\n**Codex CLI matured marketplace plugin installation and Realtime v2 defaults.** Releases `0.115.0`-`0.121.0` shipped `codex marketplace add` for installing plugins from GitHub, git URLs, local directories, and JSON feeds. TUI gained `Ctrl+R` reverse history search and slash-command recall. Memory controls, deletion, and cleanup landed alongside expanded MCP App support—resource reads, tool-call metadata, server-driven elicitations, and file uploads. Realtime v2 defaulted to WebRTC transport with configurable voice selection and native TUI media support; background agent progress streaming added. Windows sandbox enforced proxy-only networking; custom model providers now fetch short-lived bearer tokens dynamically.\n\n**Codex app expanded to in-app browser, computer use, and skills integration.** Native browser for local and public pages, computer use on macOS/Windows (unavailable in EEA/UK/Switzerland), skills accessible via `@` menu, fork-from-earlier-message support, and `/model` and `/reasoning` slash commands shipped. Thread search, archive actions, and settings sync between app and VS Code extension landed. April 7 deadline for legacy model removal (`gpt-5.2-codex`, `gpt-5.1`-variants, `gpt-5`); full removal April 14.\n\n**GPT-5.4 mini and nano released alongside specialized models.** `gpt-5.4-mini` and `gpt-5.4-nano` became available as faster, lower-cost alternatives (v6.30.0+, v3.29.0+, v2.29.0+). **GPT-Rosalind** launched for life sciences and drug discovery. **GPT-5.4-Cyber** expanded to vetted defenders through Trusted Access. Cloudflare Agent Cloud and enterprise partners deployed models for agentic workflows.","releaseCount":236,"lastContributingReleaseAt":"2026-04-16T17:42:47.000Z","generatedAt":"2026-04-08T13:00:15.799Z","updatedAt":"2026-04-17T21:12:15.137Z"},"knowledgePage":{"scope":"org","content":"**Agents SDK shipped native sandbox execution and expanded Codex with in-app browser, computer use, and marketplace plugins.**\n\n**Go, Node, and Python SDKs aligned on message execution tracking and ephemeral token security.** Go (`v3.32.0`), Node (`v6.34.0`), and Python (`v2.32.0`) added `phase` field to conversation messages for pipeline visibility into execution stages. All three implemented **short-lived token support** to reduce credential exposure in temporary sessions. Go and Node formalized `OAuthErrorCode` types; Go and Python added `detail` to `InputFileContent`. Node (`v6.33.0`-`v6.34.0`) extended async iterators and stream support to WebSocket classes. Python added WebSocket reconnection handlers and raw data transmission.\n\n**Codex CLI matured marketplace plugin installation and Realtime v2 defaults.** Releases `0.115.0`-`0.121.0` shipped `codex marketplace add` for installing plugins from GitHub, git URLs, local directories, and JSON feeds. TUI gained `Ctrl+R` reverse history search and slash-command recall. Memory controls, deletion, and cleanup landed alongside expanded MCP App support—resource reads, tool-call metadata, server-driven elicitations, and file uploads. Realtime v2 defaulted to WebRTC transport with configurable voice selection and native TUI media support; background agent progress streaming added. Windows sandbox enforced proxy-only networking; custom model providers now fetch short-lived bearer tokens dynamically.\n\n**Codex app expanded to in-app browser, computer use, and skills integration.** Native browser for local and public pages, computer use on macOS/Windows (unavailable in EEA/UK/Switzerland), skills accessible via `@` menu, fork-from-earlier-message support, and `/model` and `/reasoning` slash commands shipped. Thread search, archive actions, and settings sync between app and VS Code extension landed. April 7 deadline for legacy model removal (`gpt-5.2-codex`, `gpt-5.1`-variants, `gpt-5`); full removal April 14.\n\n**GPT-5.4 mini and nano released alongside specialized models.** `gpt-5.4-mini` and `gpt-5.4-nano` became available as faster, lower-cost alternatives (v6.30.0+, v3.29.0+, v2.29.0+). **GPT-Rosalind** launched for life sciences and drug discovery. **GPT-5.4-Cyber** expanded to vetted defenders through Trusted Access. Cloudflare Agent Cloud and enterprise partners deployed models for agentic workflows.","releaseCount":236,"lastContributingReleaseAt":"2026-04-16T17:42:47.000Z","generatedAt":"2026-04-08T13:00:15.799Z","updatedAt":"2026-04-17T21:12:15.137Z"},"playbook":{"scope":"playbook","content":"# OpenAI — Playbook\n\n> Agent reference for fetching and maintaining **OpenAI** (`openai`) changelog sources.\n\n**7** active sources · **4** products: ChatGPT, GPT API, Codex, API · domain: openai.com\n\n## Sources\n\n| Name | ID | Type | URL | Product | Last Fetched |\n|------|-----|------|-----|---------|--------------|\n| Product Releases | `src_9GF-s3jI4ndAEnclz5Yku` | feed | https://openai.com/news/product-releases/ | — | Apr 18 |\n| Python SDK | `src_jsbnrYBzcKUcWTVRDsop2` | github | https://github.com/openai/openai-python | — | Apr 18 |\n| Node SDK | `src_E8SOv7oaCXxshXd3BtEii` | github | https://github.com/openai/openai-node | — | Apr 18 |\n| .NET SDK | `src_8tt50oaCp5HlX5LowU2Cb` | github · low | https://github.com/openai/openai-dotnet | — | Apr 18 |\n| Go SDK | `src_Mqu8-CHN7z7Mii2OjcKKX` | github | https://github.com/openai/openai-go | — | Apr 18 |\n| API Changelog | `src_lUUecdtTa2GdtjHTQGpjm` | feed | https://developers.openai.com/api/docs/changelog | API | Apr 11 |\n| Codex Changelog | `src_5o4Llv42g_j-oLZvOjA9p` | feed | https://developers.openai.com/codex/changelog | Codex | Apr 16 |\n\n## Agent Notes\n\n### Fetch instructions\n- **Product Releases feed** (openai-news): Backed by openai.com/news/rss.xml. The feed is noisy — it includes academy tutorials, research papers, policy posts, and corporate news alongside actual product releases. The canonical URL is openai.com/news/product-releases/ for human consumers. When processing fetched entries, include only: model launches, API updates, ChatGPT feature announcements, SDK releases, platform changes, product launches (Codex, Operator, Sora). Suppress entries with /academy/ URLs, research papers, policy statements, partnership-only announcements, hiring posts, and corporate news without a product component.\n- **Feed sources** (openai-api-changelog, openai-codex-changelog): High-signal, canonical sources for API and Codex product changes. Structured and reliable — no filtering needed.\n- **GitHub SDKs** (openai-python-sdk, openai-node-sdk, openai-go-sdk, openai-net-sdk): Standard GitHub release extraction. Conventional semver releases with changelogs. No filtering needed.\n\n### Traps\n- **Academy noise in Product Releases feed**: The RSS feed mixes educational content (/academy/* URLs) with product announcements. These are not product releases — suppress aggressively.\n- **No dedicated ChatGPT feed**: ChatGPT product changes only surface through the Product Releases feed or the API changelog. No single source covers ChatGPT comprehensively.\n\n### Coverage\n- **API product**: openai-api-changelog (primary) + four GitHub SDK repos (supporting).\n- **Codex product**: openai-codex-changelog (primary).\n- **ChatGPT product**: No dedicated source — changes appear in the Product Releases feed or API changelog if API-adjacent.\n- **GPT API product**: Overlaps with openai-api-changelog.\n- **SDK sources**: Org-level, spanning API and GPT API products.\n\n### Release cadence\nOpenAI ships continuously. The API changelog and Codex changelog update multiple times per week. SDK releases cluster around API changes. The Product Releases feed publishes several times weekly but most entries are non-product content that should be suppressed.\n","updatedAt":"2026-04-18T20:42:50.722Z"}}