11:38:05 RRSAgent has joined #webmachinelearning 11:38:09 logging to https://www.w3.org/2026/10/30-webmachinelearning-irc 11:38:09 RRSAgent, make logs Public 11:38:10 please title this meeting ("meeting: ..."), anssik 11:38:11 Meeting: WebML CG F2F TPAC 2026 – 30 October 2026 11:38:12 Chair: Anssi 11:38:15 Agenda: https://github.com/webmachinelearning/meetings/issues/39 11:38:19 Scribe: Anssi 11:38:23 scribeNick: anssik 11:38:31 RRSAgent, draft minutes 11:38:32 I have made the request to generate https://www.w3.org/2026/10/30-webmachinelearning-minutes.html anssik 11:38:37 Topic: Built-in AI APIs overview 11:38:49 Present+ Mike Wasserman 11:38:55 ota has joined #webmachinelearning 11:39:38 present+ 11:39:40 Present+ Reilly_Grant 11:39:43 present+ 11:39:44 Present+ Anssi_Kostiainen 11:39:46 present+ Gaurang Tandon 11:39:48 present+ ali spivak 11:39:49 present+ 11:39:55 present+ Marijn_Kruisselbrink 11:39:58 Present+ Mathieu Giroux-Huppe 11:40:14 Anssi: any implementation status updates from Chrome, Edge or others? 11:40:22 present + 11:40:27 ... Kenji has kept Awesome Built-in AI APIs repo updated for applications & demos that make use of these APIs: 11:40:31 -> https://github.com/webmachinelearning/awesome-built-in-ai 11:40:34 sushraja has joined #webmachinelearning 11:40:43 q+ reillyg 11:40:44 ack reillyg 11:40:56 q+ 11:41:43 ack sushraja 11:42:24 RRSAgent, draft minutes 11:42:25 I have made the request to generate https://www.w3.org/2026/10/30-webmachinelearning-minutes.html anssik 11:42:48 q+ reillyg 11:42:49 ack reillyg 11:43:13 Dingwei has joined #webmachinelearning 11:45:30 Topic: Prompt API 11:45:38 gb, this is webmachinelearning/prompt-api 11:45:38 anssik, OK. 11:45:47 Subtopic: Model selection and availability 11:45:56 Anssi: issue #169 11:45:57 https://github.com/webmachinelearning/prompt-api/issues/169 -> Issue 169 [Tag Review] - Model selection and availability (by etiennenoel) [tag-tracker] 11:46:01 ... this issue discussed Mozilla's and in part also WebKit's and TAG's concerns 11:46:26 ... we made a resolution that address these concerns: "Prompt API specification to require conformant implementations to use a permissively licensed model provided either by the implementation or the underlying platform. As further work, determine what criteria makes sense exactly for such models. (issue #169)" 11:46:30 -> 11 June minutes https://www.w3.org/2026/06/11-webmachinelearning-minutes.html#7b62 11:46:39 Anssi: any questions on how to interpret or implement this resolution? 11:46:41 q+ reillyg 11:46:42 ack reillyg 11:47:01 mklimenko has joined #webmachinelearning 11:48:06 q+ sushraja 11:48:08 ack sushraja 11:49:54 q+ Sohan_Jyoti 11:50:04 ack Sohan_Jyoti 11:50:15 q+ 11:50:23 q+ reillyg 11:50:26 ack reillyg 11:52:23 q+ sushraja 11:52:24 ack sushraja 11:53:36 q+ Chris 11:53:39 ack Chris 11:53:43 sushraja has joined #webmachinelearning 11:53:57 q+ RRSAgent 11:54:00 q+ reillyg 11:54:03 ack reillyg 11:54:08 ack RRSAgent 11:54:21 Present+ Sushanth_Rajasankar 11:56:49 present+ 11:56:59 q+ reillyg 11:57:01 ack reillyg 11:58:52 q+ MarkusT 11:58:56 ack MarkusT 11:58:59 q+ 12:00:08 q- 12:00:34 q+ 12:01:16 Anssi: in response to MarkusT's comment, there's an issue for letting the user choose a model, see issue #8 12:01:16 https://github.com/webmachinelearning/prompt-api/issues/8 -> Issue 8 Choose model (by niutech) 12:01:25 ack reillyg 12:02:07 Reilly: the ecosystem of models is evolving, COS API takes an approach that is opposite to built-in AI APIs 12:02:15 q+ 12:02:19 ... middleground for task-focused APIs, users can pick a model 12:02:41 ... interop one is important, benefits to browser vendors to optimize a model to run on the device 12:02:57 ... developer choice means they lose control over model performance 12:03:06 ... a tradeoff to take 12:03:37 q+ MarkusT 12:03:39 ack MarkusT 12:04:19 q+ reillyg 12:04:20 ack reillyg 12:05:05 ack sushraja 12:06:24 Present+ Markus_Tavenrath 12:06:43 Topic: Message format for tool calling, thinking, and multimodality 12:06:48 mtavenrath has joined #webmachinelearning 12:07:05 q+ reillyg 12:07:06 ack reillyg 12:07:40 Anssi: a number of features under development, message format for tool calling, thinking, and multimodality #209 12:07:40 sampling modes #203, thinking effort #222, and fastest mode #219 12:07:41 https://github.com/webmachinelearning/prompt-api/issues/203 -> Issue 203 Proposal: Introduce Categorical Sampling Modes for the Prompt API (by isaacahouma) [enhancement] [interop] 12:07:41 https://github.com/webmachinelearning/prompt-api/issues/209 -> Issue 209 Extend the Message format to support tool calling, extended thinking, and multimodality (by nico-martin) 12:07:41 https://github.com/webmachinelearning/prompt-api/issues/222 -> CLOSED Issue 222 Thinking effort API - includeThoughts (by sushraja-msft) 12:07:42 https://github.com/webmachinelearning/prompt-api/issues/219 -> Issue 219 Consider adding a "fastest" LanguageModelSamplingMode value (by isaacahouma) 12:08:00 q+ 12:08:05 s/Topic: Message format for tool calling, thinking, and multimodality/Topic: New features under development 12:08:21 RRSAgent, draft minutes 12:08:22 I have made the request to generate https://www.w3.org/2026/10/30-webmachinelearning-minutes.html anssik 12:08:43 q+ 12:08:50 s/Topic: New features under development/Subtopic: New features under development 12:08:51 RRSAgent, draft minutes 12:08:53 I have made the request to generate https://www.w3.org/2026/10/30-webmachinelearning-minutes.html anssik 12:10:00 ack sushraja 12:11:10 ack mklimenko 12:11:35 q+ reillyg 12:11:36 ack reillyg 12:12:26 q? 12:12:35 q+ reillyg 12:12:37 ack reillyg 12:13:06 Present+ Ehsan_Toreini 12:13:24 Present+ Samuel_Dallstream 12:13:54 q+ 12:14:02 ack isaac 12:14:27 q+ sushraja 12:14:28 ack sushraja 12:15:07 Present+ Andrew_Wafaa 12:16:17 Topic: Writing Assistance APIs 12:16:19 q+ reillyg 12:16:20 ack reillyg 12:16:51 gb, this is webmachinelearning/writing-assistance-apis 12:16:51 anssik, OK. 12:16:53 q+ 12:17:44 ack sushraja 12:18:56 s/Topic: Writing Assistance APIs/Topic: Agenda bashing 12:19:18 q+ reillyg 12:19:19 ack reillyg 12:19:46 q+ 12:20:20 q- 12:20:23 Topic: Translator and Language Detector APIs 12:20:25 q+ sushraja 12:20:27 ack sushraja 12:20:46 Subtopic: Batched translation 12:20:49 Anssi: issue #85 12:20:50 Issue 85 not found 12:21:11 gb, this is webmachinelearning/translation-api 12:21:11 anssik, OK. 12:21:31 Sushant: batched translation is on pause until we learn more 12:21:53 s/batched translation/batched translation #85 12:21:55 q+ reillyg 12:21:59 ack reillyg 12:22:05 RRSAgent, draft minutes 12:22:07 I have made the request to generate https://www.w3.org/2026/10/30-webmachinelearning-minutes.html anssik 12:22:10 q+ 12:22:35 q+ ehsan 12:22:37 ack ehsan 12:22:57 q+ reillyg 12:22:58 ack reillyg 12:23:20 ack mklimenko 12:23:44 q+ 12:24:00 +1 on shushraja on the status on translation api and potential to move it to wg 12:24:28 q+ reillyg 12:24:34 ack reillyg 12:25:03 ack mtavenrath 12:26:39 Topic: Proposals for new incubations 12:26:45 gb, this is webmachinelearning/proposals 12:26:45 anssik, OK. 12:26:54 Subtopic: Decisions API 12:27:00 Anssi: Decisions API #20 12:27:00 https://github.com/webmachinelearning/proposals/issues/20 -> Issue 20 Decision Models (by AdamSobieski) [Agenda+] 12:27:05 -> Explainer for the Decisions API https://github.com/explainers-by-googlers/decisions-api 12:27:13 q+ reillyg 12:27:17 ack reillyg 12:29:27 [ Decisions API demo: Token Sushi ] 12:30:13 Subtopic: Semantic Embedder API 12:30:17 -> Semantic Embedder API https://github.com/explainers-by-googlers/semantic-embedder-api 12:30:27 q+ reillyg 12:30:28 ack reillyg 12:30:48 RRSAgent, draft minutes 12:30:49 I have made the request to generate https://www.w3.org/2026/10/30-webmachinelearning-minutes.html anssik 12:31:00 q+ 12:31:59 ack toreini 12:32:43 sushraja has joined #webmachinelearning 12:33:10 -> https://github.com/explainers-by-googlers/vector-store-api 12:34:32 I think perhaps the resolution for the model calcification should be added to these two new explainers. 12:35:18 +1 12:35:21 RESOLUTION: Request TAG feedbac 12:35:28 RESOLUTION: Request TAG feedback on Semantic Embedder API v Built-in Vector Store API. 12:35:36 s/RESOLUTION: Request TAG feedbac// 12:36:03 RRSAgent, draft minutes 12:36:04 I have made the request to generate https://www.w3.org/2026/10/30-webmachinelearning-minutes.html anssik 12:36:52 s/Request TAG feedbac/Request TAG feedback on Semantic Embedder API v Built-in Vector Store API./ 12:36:58 RRSAgent, draft minutes 12:36:59 I have made the request to generate https://www.w3.org/2026/10/30-webmachinelearning-minutes.html anssik 12:38:36 q+ Yunhao_Luo 12:38:39 Present+ Yunhao_Luo 12:38:50 q+ reillyg 12:38:52 ack reillyg 12:39:18 s/k on Semantic Embedder API v Built-in Vector Store API.// 12:39:28 RRSAgent, draft minutes 12:39:30 I have made the request to generate https://www.w3.org/2026/10/30-webmachinelearning-minutes.html anssik 12:39:49 q+ Yunhao_Luo 12:39:53 ack Yunhao_Luo 12:40:32 Present+ Sohum_Chatterjee 12:41:45 q+ 12:41:49 q+ 12:42:43 ack sushraja 12:43:08 ack toreini 12:43:33 q+ reillyg 12:43:35 ack reillyg 12:45:16 Topic: Wrap up 12:45:21 Anssi: thank you for joining this Built-in AI APIs F2F meeting, it was a blast 12:45:27 ... we'll share the minutes after a post-processing step and human review 12:45:36 RRSAgent, draft minutes 12:45:37 I have made the request to generate https://www.w3.org/2026/10/30-webmachinelearning-minutes.html anssik 13:14:56 toreini has joined #webmachinelearning 14:09:39 toreini has joined #webmachinelearning 15:02:52 Zakim has left #webmachinelearning 15:44:54 toreini has joined #webmachinelearning 15:46:44 Penny has joined #webmachinelearning 16:05:43 Penny has joined #webmachinelearning 16:10:25 Penny has joined #webmachinelearning 16:25:01 Penny has joined #webmachinelearning 16:33:02 Penny has joined #webmachinelearning 17:21:17 Penny has joined #webmachinelearning 17:35:03 Penny has joined #webmachinelearning 17:56:38 Penny has joined #webmachinelearning 19:52:35 Penny has joined #webmachinelearning 19:53:08 Penny has joined #webmachinelearning