Fix mount-time settings clobbering in website chat

[posthog]

Changes

Fixes four defects in the posthog.com chat that a customer reported as "pretty much unusable." All four are in src/hooks/useChat.tsx / src/hooks/useInkeepSettings.ts.

Why: every visitor and prospect can open this chat. The defects sit on the main path of the feature, and cost the user correct answers and the ability to restore a past conversation.

• Lost page context and example questions. Three effects each did setAiChatSettings({ ...aiChatSettings, ... }) with aiChatSettings missing from their dependency arrays. On mount all three read the same initial object, so only the last write survived and prompts and exampleQuestions were dropped — the chat answered without knowing what page you were on. Each write is now a functional update, so nothing clobbers anything.
• Dropped chatId and theme. The same stale-spread crossed hooks: useInkeepSettings wrote colorMode, then useChat overwrote it, and a later change could wipe chatId, so restoring a conversation from history did not load it. Both writes are now functional merges.
• Frozen onEvent closure. The Inkeep onEvent callback was installed once in a [] effect, so it held mount-time firstResponse / hasFirstResponse forever. It is now reinstalled when the callback identity changes.
• Leaked shadow-root listeners. Each assistant_answer_displayed added click listeners to every link without removing them. One delegated listener now handles all links, and it is removed on unmount.

Also adds two posthog.capture calls (chat question submitted, chat answer displayed) — the surface had no instrumentation, so the impact could not be sized in data.

No visual change: the fixes restore correct behavior on an unchanged UI. The chat renders through a live external (Inkeep) service that needs an API key, so this was verified by code reasoning and static checks (prettier, type check), not a live reproduction — consistent with the report, which notes the surface is uninstrumented and could not be sized in data.

Checklist

• I've read the docs and/or content style guides.
• Words are spelled using American English
• Use relative URLs for internal links
• I've checked the pages added or changed in the Vercel preview build
• If I moved a page, I added a redirect in vercel.json

---

Created with PostHog Desktop from this inbox report.

[posthog]

🦔 PostHog Review reviewed this pull request

Found 0 must fix, 4 should fix, 0 consider.

Published 4 findings (view the review).

Resolved comments: 3 fixed

[github-actions]

Deploy preview

Status	Details	Updated (UTC)
🟢 Ready	View preview	Aug 30, 2026 01:07PM

[github-actions]

Bundle report

Total JS (gzip)

8.15 MiB (+0.1 KiB / +0.0%)

Eager graph (modules shipped in each entrypoint's initial chunks)

Entrypoint	Eager size	Budget	Modules
✅ app	16.91 MiB (+2.5 KiB / +0.0%)	report-only	2037

Largest modules in the app closure

Module	Size
./src/data/mcp-tools.json	1054.5 KiB
css ./node_modules/.pnpm/css-loader@5.2.7_webpack@5.101.3/node_modules/css-loader/dist/cjs.js??ruleSet[1].rules[8].oneOf[1].use[1]!./node_modules/.pnpm/postcss-loader@4.3.0_postcss@8.5.6_webpack@5.101.3/node_modules/postcss-loader/dist/cjs.js??ruleSet[1].rules[8].oneOf[1].use[2]!./src/styles/global.css	754.5 KiB
./src/components/Stickers/Stickers.tsx	696.4 KiB
./node_modules/.pnpm/@radix-ui+react-icons@1.3.2_react@18.3.1/node_modules/@radix-ui/react-icons/dist/react-icons.esm.js	481.4 KiB
./node_modules/.pnpm/rehype-raw@7.0.0/node_modules/rehype-raw/lib/index.js + 29 modules	395.1 KiB
./src/hooks/useCustomers.tsx + 55 modules	370.0 KiB
./node_modules/.pnpm/@posthog+icons@0.36.6_react-dom@18.3.1_react@18.3.1__react@18.3.1/node_modules/@posthog/icons/dist/posthog-icons.es.js	354.8 KiB
./node_modules/.pnpm/react-markdown@8.0.7_@types+react@16.14.66_react@18.3.1/node_modules/react-markdown/lib/react-markdown.js + 88 modules	351.4 KiB
./src/components/ProductComparisonTable/index.tsx + 126 modules	301.7 KiB
./node_modules/.pnpm/cloudinary-core@2.14.0_lodash@4.17.21/node_modules/cloudinary-core/cloudinary-core.js	281.9 KiB
./src/components/SearchUI/index.tsx + 87 modules	273.0 KiB
./node_modules/.pnpm/@posthog+brand@0.8.0_react@18.3.1/node_modules/@posthog/brand/dist/generated/hoggies/svg/magnifying-glass.mjs	254.7 KiB
./node_modules/.pnpm/framer-motion@10.18.0_react-dom@18.3.1_react@18.3.1__react@18.3.1/node_modules/framer-motion/dist/es/render/dom/motion.mjs + 109 modules	253.9 KiB
./node_modules/.pnpm/d3@7.9.0/node_modules/d3/src/index.js + 208 modules	247.4 KiB
./src/components/Pricing/PricingSlider/Slider.tsx + 87 modules	240.1 KiB

---

Eager-graph budgets are report-only until a baseline is established. Sizes are gzip of public/**/*.js; eager size is webpack module source bytes for the modules actually shipped in the entrypoint's initial chunks (post-tree-shake).

[posthog]

PostHog Review alpha 🦔 If you find any issues helpful - please reply "valid", "invalid", etc., for evaluation purposes 🙏

[posthog]

PostHog Review

Found 4 should fix.

Other findings (outside the changed lines)

Valid issues on this PR's files that sit on lines GitHub won't let us comment on inline.

Question analytics omit all follow-up questions

Priority: should_fix | File: src/hooks/useChat.tsx:82-86 | Category: bug

Why we think it's a valid issue

• Checked: the logEventCallback body at src/hooks/useChat.tsx:80-124, the firstResponse state and where it resets, the effect that reinstalls onEvent at src/hooks/useChat.tsx:179-186, the provider key at src/hooks/useChat.tsx:262, and every other posthog.capture call site in src/.
• Found: src/hooks/useChat.tsx:82 guards the whole block with !firstResponse, and the capture at line 86 sits inside that block. Line 83 sets firstResponse from the first user message, so the guard is false for the rest of the conversation. The capture at src/hooks/useChat.tsx:93-94 carries no equivalent guard, and Inkeep raises assistant_answer_displayed once per answer.
• Found: this PR is what makes the guard effective. Before the change, onEvent held a mount-time closure, so firstResponse stayed null and line 86 would have fired on each message. The new effect at src/hooks/useChat.tsx:179-186 reinstalls onEvent when logEventCallback changes, so the callback now reads the current firstResponse. The one-per-conversation behavior arrives with this change.
• Found: no other call site captures a chat question. The only other chat captures are Opened MaxAI chat at src/components/AskAIInput/index.tsx:29 and src/components/AskMax/index.tsx:47, which record a chat open, not a question.
• Found: src/hooks/useChat.tsx:262 keys ChatProvider on the chat id or path, so firstResponse resets only when a new conversation mounts. The event therefore counts conversations, while its name states chat question submitted.
• Impact: every follow-up question in a conversation produces a chat answer displayed event with no matching chat question submitted event. The two counts diverge by the number of follow-ups, so a question-to-answer comparison is wrong and total question volume reads low. Multi-turn use stays invisible in the data. The PR adds these events to size the chat in data, so the wrong count works against the reason for the change. The fix moves one line out of the guard and keeps setFirstResponse inside it.

Issue description

The capture call is inside the !firstResponse condition. It records only the first user message. The answer event records every answer. Follow-up questions therefore create unmatched answer events and incorrect conversion counts.

Suggested fix

Capture every user_message_submitted event. Keep only setFirstResponse(...) inside the if (!firstResponse) condition.