14:58:59 RRSAgent has joined #webmachinelearning 14:59:03 logging to https://www.w3.org/2026/06/25-webmachinelearning-irc 14:59:03 RRSAgent, make logs Public 14:59:04 please title this meeting ("meeting: ..."), anssik 14:59:04 Meeting: WebML CG Teleconference – 25 June 2026 14:59:15 Chair: Anssi 15:01:14 Present+ Juan_Fernandez 15:01:19 Agenda: https://github.com/webmachinelearning/meetings/blob/main/telcons/2026-06-25-cg-agenda.md 15:01:22 Scribe: Anssi 15:01:28 ScribeNick: anssik 15:01:45 Present+ Anssi_Kostiainen 15:01:59 msw has joined #webmachinelearning 15:02:06 Present+ Reilly_Grant 15:02:15 Present+ Benjamin_VanderSloot 15:02:20 Present+ Dominic_Farolino 15:02:23 RRSAgent, draft minutes 15:02:25 I have made the request to generate https://www.w3.org/2026/06/25-webmachinelearning-minutes.html anssik 15:02:52 Anssi: as a reminder, we'll use IRC-based queue management in this meeting: 15:02:57 -> https://www.w3.org/guide/meetings/zakim.html#speakerqueue 15:03:06 Anssi: to suggest agenda topics, use Agenda+ label, e.g.: 15:03:19 -> Agenda+ https://github.com/webmachinelearning/webmcp/labels/Agenda+ 15:03:23 Victor has joined #webmachinelearning 15:03:32 Anssi: please welcome our latest new participants 15:03:32 ... Erik Nordin from Mozilla 15:03:38 ... Juan Fernandez from Shopware 15:03:45 kush has joined #webmachinelearning 15:04:27 Topic: Announcements 15:04:32 Subtopic: TPAC 2026 schedule 15:04:37 Anssi: TPAC 2026 is in Dublin, Ireland on 26-30 October 2026 15:04:54 ... the WebML WG is planning to meet on Mon 26 October 2026 and the WebML CG on Tue 27 October 2026 15:04:58 ... given the two groups have good overlap in both participation and expertise, we make this a joint F2F meeting 26-27 October 2026 similarly to TPAC 2025 15:05:00 bvandersloot has joined #webmachinelearning 15:05:11 ... Monday primarily for WebNN API topics, and Tuesday for WebMCP, Prompt API and other built-in AI APIs 15:05:11 ... I have opened a F2F GH issue to solicit ideas for the F2F agenda: 15:05:12 -> https://github.com/webmachinelearning/meetings/issues/39 15:05:13 https://github.com/webmachinelearning/meetings/issues/39 -> Issue 39 WebML WG/CG F2F Agenda - TPAC 2026 (Dublin, Ireland) (by anssiko) 15:05:24 Anssi: as is practical, the detailed agenda will be worked out closer to the meeting, but I wanted to give everyone a heads-up and an opportunity to provide input early 15:05:28 ... any questions or suggestions? 15:05:38 q? 15:05:45 Topic: Prompt API 15:05:49 gb, this is webmachinelearning/prompt-api 15:05:49 anssik, OK. 15:05:53 Subtopic: Sampling parameters 15:05:57 Anssi: issue #203 15:05:58 https://github.com/webmachinelearning/prompt-api/issues/203 -> Issue 203 Proposal: Introduce Categorical Sampling Modes for the Prompt API (by isaacahouma) [enhancement] [interop] [Agenda+] 15:06:12 ... we have early developer feedback result on how developers are interacting with sampling parameters 15:06:18 ... e.g. temperature and topK 15:06:50 Anssi: Kenji shares "45% of respondents stated they are actively tuning sampling parameters in their workflows." 15:07:20 Reilly: we found out there's a reasonably large group that can migrate to the pre-defined sampling parameters 15:07:49 Ehsan has joined #webmachinelearning 15:07:53 MikeWa: we are collecting feedback from developers and want specific examples to understand the gaps 15:08:23 domfarolino has joined #webmachinelearning 15:09:13 MikeWa: to close the gaps we have for behaviors, we know the developers want a mode that has lowers predictability 15:09:18 fr has joined #webmachinelearning 15:09:52 ... summarization or information extraction use cases considered, adjusting parameters of the 5 modes we have and seek further input and concrete examples from developers 15:10:03 ... we may want to add 1-2 additional modes 15:10:29 ... happy to report when we get additional information and data 15:12:15 gurusingh has joined #webmachinelearning 15:12:27 q? 15:13:01 lgombos has joined #webmachinelearning 15:13:17 BenGreenstein has joined #webmachinelearning 15:13:30 Anssi: we'll await further data from developers and will revisit this issue 15:13:55 Topic: WebMCP 15:13:58 present+ Laszlo_Gombos 15:13:58 gb, this is webmachinelearning/webmcp 15:13:59 anssik, OK. 15:14:05 Subtopic: Wide review 15:14:18 Anssi: we are ready to initiate the review process with the TAG, Privacy WG and Security IG 15:14:25 ... thank you all for your work in preparing the review materials, now approved by the group: 15:14:29 -> Explainer https://github.com/webmachinelearning/webmcp/blob/main/README.md 15:14:33 -> S&P Considerations https://webmachinelearning.github.io/webmcp/#security-privacy 15:14:36 -> S&P Questionnaire https://github.com/webmachinelearning/webmcp/blob/main/security-privacy-questionnaire.md 15:15:33 Anssi: I should note the Security IG already discussed WebMCP in their 9 June 2026 meeting and they are supportive of the review process, awaiting the formal review request from the CG, thanks Johann for joining that meeting 15:15:35 Winston has joined #webmachinelearning 15:15:38 -> Security IG 2026-06-09 minutes https://github.com/w3c/securityig/blob/main/meetings/2026/2026-06-09_minutes.md#webmcp-early-review-and-threat-modeling-scope 15:16:20 Anssi: as a reminder, here are the templates for the review requests for the TAG, Privacy WG and Security IG: 15:16:26 -> TAG review request ("Early design/incubation review") https://github.com/w3ctag/design-reviews/issues/new?template=000-incubation-review.yaml 15:16:29 -> Privacy review request template ("Request review for another type of spec") https://github.com/w3cping/privacy-request/issues/new?template=request-review-for-another-type-of-spec.md 15:16:33 -> Security review request template ("Request review for another type of spec") https://github.com/w3c/security-request/issues/new?template=request-review-for-another-type-of-spec.md 15:18:24 +1 15:18:32 RESOLUTION: Initiate wide review of WebMCP with the TAG, Privacy WG and Security IG. 15:18:36 +1 15:18:56 Subtopic: WG transition 15:19:05 Anssi: we are in process of transitioning WebMCP from the CG to the WG 15:19:18 ... I discussed this matter with the WebML WG last week and shared the charter change proposal with the group: 15:19:19 I will ask Brandon if he wants to participate in the review as well 15:19:22 -> https://www.w3.org/2026/06/18-webmachinelearning-minutes.html#0b72 15:19:44 Anssi: we are expected to seek review from the TAG, Security IG, Privacy WG, a11y/ARIA WG to inform this transition 15:19:58 ... by initiating the wide review process, we're already taking the right steps toward this transition 15:20:09 ... it is a normal and a healthy part of the process to see an active discussion when work is transitioning to a WG 15:20:21 ... this lively discussion emphasizes the importance of the work 15:20:33 q+ 15:20:46 ... each W3C Member has one vote to ensure the entire W3C membership is represented in the decision-making process 15:20:52 ack domfarolino 15:22:07 Dominic: when we filed the standards position with Apple, the early feedback was to move away from the current group structure 15:22:23 ack domfarolino 15:23:51 Saron: I haven't been part of this discussion, we see the problem as an HTML specific problem 15:24:58 Dominic: would be great to talk to Apple folks about their feedback 15:25:10 q? 15:27:14 BenV: Mozilla doesn't have a strong opinion, it makes sense to advance from the CG stage 15:28:05 Laszlo: we have a similar opinion, we want to see this advance to a WG 15:28:20 q? 15:28:40 Subtopic: Framework considerations 15:28:44 Anssi: issue #199 15:28:45 https://github.com/webmachinelearning/webmcp/issues/199 -> Issue 199 Frameworks and WebMCP Meta Issue (by sdras) [Agenda+] 15:29:44 ... Sarah opened a meta issue for framework considerations for WebMCP, thank you for helping the group understand the framework landscape better 15:29:50 ... let's first build understanding of what are typical frameworks in this context, how frameworks interact with WebMCP, and what the implications are for WebMCP design 15:29:55 ... after this stage setting, let's discuss the specific framework considerations in separate topics I scheduled for this meeting 15:30:12 Dominic: in broad strokes, there are ways how WebMCP can play nicer with frameworks 15:30:58 ... Sarah is doing the work to identify opportunities how that can happen 15:32:38 ... we're thinking how the first sub-issue would work, we could have tools with closure that become stale 15:33:52 ... we're trying to understand what is the concrete issue with stale closures, so we could derive a WebMCP solution 15:34:22 ... also want to understand if this can be solved in the userland 15:35:19 for "A pending/unavailable tool state" consideration and "Persistence across routes/views/pages" we want to have site be able to register what tools are available to a page before the script has executed and registered the tools 15:36:03 ... we're exploring declarative mechanisms for this 15:36:55 ... when we talk about tool persistence, navigation requires registration trashing in model context, possibly 15:37:24 ... tool execution persistence, same origin navigations, two angles or persistence 15:37:53 ... what the concrete issues for the agent here are, "A pending/unavailable tool state" is more relevant from clients I think 15:37:55 q? 15:38:48 q+ 15:39:14 ack bvandersloot 15:40:09 BenV: I think the couple of versions of persistence are interesting, thinking of multi-page web pages, context manager would grow when you navigate the site, dropping tool definitions would not be good 15:40:16 ... I'm trying to keep that design in mind 15:40:55 Dominic: two things, framework specific state management shared by React and Vue 15:41:22 ... secondly, persistence considerations that are also important outside framework considerations 15:44:14 Juan: we use Vue and try integrate it with WebMCP, I can share my feedback in the meta issue #199 15:44:14 https://github.com/webmachinelearning/webmcp/issues/199 -> Issue 199 Frameworks and WebMCP Meta Issue (by sdras) [Agenda+] 15:44:46 +1 15:45:01 +1 15:45:04 RESOLUTION: Investigate state management and persistence requirements for frameworks (issue #199) 15:45:36 Subtopic: Dynamic tool definitions 15:45:42 Anssi: issue #167 15:45:43 https://github.com/webmachinelearning/webmcp/issues/167 -> Issue 167 Proposal: Dynamic Tool Definitions in WebMCP (by AlpineJosh) [Agenda+] 15:45:53 ... a proposal to let tool definitions respond to state 15:46:10 ... per this issue author, registerTool() assumes inputSchema and callability to be fixed at registration time 15:46:19 ... proposal is to lazily resolve at enumeration time 15:46:31 ... and add a disabled flag via updateTool() to indicate that a tool is not callable at the moment 15:46:48 ... example: a music app's removeFromQueue tool is meaningless when queue is empty, should be enabled only when queue is non-empty 15:47:09 Anssi: Dominic responded unregister and re-register the tool yields the same effect 15:47:15 ... unregistering happens via an AbortController signal, maybe the ergonomics of that is not great 15:47:18 ... and could be improved with updateTool()? 15:47:46 Dominic: my opinion is we can add updateTool(), but I'd like to hear more from model providers 15:48:02 ... does this solve a problem that unregister and register wouldn't solve 15:48:39 ... some trashing is involved when unregister and register happens 15:49:08 ... Victor, do you have any insights whether trashing happens between each call? 15:49:08 q+ 15:49:09 q? 15:49:37 Dominic: with Genimi Chrome experience, I don't see this as a problem 15:50:20 q+ 15:50:22 ack Victor 15:50:41 Victor: about updateTool() I'm curious how to affects the model side 15:50:51 ... I see this as developer ergonomics consideration 15:51:01 ... the model should be a consideration when we're mid-flight 15:51:29 ... allowing a base condition, might add more complications how the tool calling happens 15:51:43 Dominic: changing the tool while execution happens, I don't think we want to support that 15:52:02 ... I consider this an ergonomics issue, also want to understand if this an agent ergonomics improvement 15:52:35 Victor: if you can update it, you can add a diff and no need to recreate the model context 15:53:24 ... it feels on the model developer this would help, but we should focus on the ergonomics improvement 15:53:48 ... I don't there's a big issue there 15:53:57 s/there's/think there's 15:54:46 Kushal: this issue is about keeping the tool name the same but updating the schema, some parts of the tool definition that are model exposed 15:55:11 ... I think from that case, the integration in Gemini Chrome, the model is given a task, I want to see what the current state of the tool looks like 15:55:36 ... after the tool has executes we'll pull the current state, updating the inputSchema could happen lazily, but not sure how that could happen 15:55:51 ... at every turn of the planning loop, all tools sets to the model 15:56:15 q+ 15:56:21 ack kush 15:56:55 Kushal: even with always on LLM looking at all available tools, I don't see how an API that let's update tool description is making this any easier 15:56:56 q? 15:57:22 Victor: because you update a small section of tool definition, you can add the diff to the context, no need to overwrite the entire cache 15:57:53 ... possible context optimization case, whole context help, previous cache not invalidated, you get to keep the caching benefits 15:58:00 Dominic: speculative 15:58:09 s/very speculative 15:58:38 Khushal: do KV cache experience apply to function calling case? 15:59:14 Reilly: want developers to cache, not gotten to function calling aspect of the API to tell if tool state changing between queries is the most efficient way 15:59:15 q? 15:59:27 ack bvandersloot 15:59:50 BenV: use case with only callback changing would be nice in a way context not altered 16:00:03 ... that structure would be a strong case for an updateTool() method 16:00:13 Khushal: the previous issue for this was about that 16:00:39 ... I support that idea 16:01:06 Dominic: we should sketch a proposal for that and table the rest 16:01:06 q? 16:01:59 proposed RESOLUTION: sketch out an API that allows updating non-model-exposed tool parts; table the rest until WebMCP consumers demonstrate a problem that demands updating model-exposed tool parts. 16:02:08 +1 16:02:40 RESOLUTION: Sketch out an API that allows updating non-model-exposed tool parts; table the rest until WebMCP consumers demonstrate a problem that demands updating model-exposed tool parts. (issue #167) 16:02:41 https://github.com/webmachinelearning/webmcp/issues/167 -> Issue 167 Proposal: Dynamic Tool Definitions in WebMCP (by AlpineJosh) [Agenda+] 16:04:13 RRSAgent, draft minutes 16:04:14 I have made the request to generate https://www.w3.org/2026/06/25-webmachinelearning-minutes.html anssik 16:24:22 s/result on/results on 16:25:09 s/lowers/lower 16:31:30 s/from clients/for clients 16:34:25 s/to affects/this affects 16:35:14 s/this an agent/this is an agent 16:36:12 s/on the model developer this would help/this would help the model developer 16:36:28 s/issue there/issue here 16:37:23 s/has executes/has executed 16:37:59 s/tools sets/tools are sent 16:38:39 s/always on/always-on 16:39:06 s/context help/context helps 16:39:44 s/speculative/very speculative 17:57:39 RRSAgent, draft minutes 17:57:40 I have made the request to generate https://www.w3.org/2026/06/25-webmachinelearning-minutes.html anssik 17:58:22 Present+ Iris_Johnson 17:58:40 Present+ Mike_Wasserman 17:59:18 Present+ Saron_Yitbarek 17:59:31 Present+ Victor_Huang 17:59:42 Present+ Guru_Singh 18:00:03 Present+ Winston_Chen 18:00:30 Present+ Khushal_Sagar 18:00:50 Present+ Neelakandan_Nc 18:01:06 Present+ Ehsan_Toreini 18:01:25 RRSAgent, draft minutes 18:01:26 I have made the request to generate https://www.w3.org/2026/06/25-webmachinelearning-minutes.html anssik