https://sonarly.com/issue/27119?type=bug
Chrome tab crashes with OOM (error code 5) because the Jotai atom family cache (`atomCache` Map) never evicts entries, and each AI chat message is stored three times (thread-level atom, per-message atom via structuredClone, Apollo InMemoryCache), causing unbounded heap growth during long conversations, rapid workspace switching, or Settings page navigation.
Fix: Three changes that address the two largest memory growth vectors in the AI chat module:
1. **Added `evictFamilyKey` to `ComponentFamilyState`** (`ComponentFamilyState.ts` type + `createAtomComponentFamilyState.ts` implementation): Exposes `atomCache.delete()` so consumers can remove atoms they no longer need. This is the missing primitive — the atomCache Map previously had no way to remove entries.
2. **Evict per-message atoms on thread switch** (`AgentChatStreamingPartsDiffSyncEffect.tsx`): Tracks message IDs seen during diff sync in a ref. When `currentAIChatThread` changes, the cleanup effect fires and calls `evictFamilyKey` for every tracked message ID, removing those atoms from the cache. This prevents the biggest accumulation vector: hundreds of per-message atoms (keyed by UUID) holding large ExtendedUIMessage payloads that were never released.
3. **Remove `structuredClone` from `useUpdateStreamingPartsWithDiff.ts`**: The deep clone was creating a full copy of every message object before storing it in the per-message atom. Since `isDeeplyEqual` already gates updates (preventing unnecessary writes), and the message reference from the thread-level atom is stable within a render cycle, the clone is unnecessary. Removing it eliminates one full copy of every message in memory.
Together these changes bound per-message atom memory to the current thread's messages (instead of all threads ever viewed), and halve the per-message memory footprint by eliminating deep clones.
**Not addressed in this PR** (requires separate, wider-scoped changes):
- Apollo InMemoryCache eviction — needs typePolicies configuration in the shared Apollo factory
- Thread-level family state atoms (10 per thread) — small scalar values, low priority
This PR:
- Breaks useAgentChatData into focused effect components (streaming,
fetch,
init, auto-scroll, diff sync)
- Splits message list into non-last (stable) + last (streaming/error) to
prevent full re-renders on each stream chunk
- Adds scroll-to-bottom button and MutationObserver-based auto-scroll on
thread switch
- Lifts loading state from context to atoms
- Adds areEqual to selector
factories
We could improve further but this sets up a robust architecture for
further refactoring.
## Messages flow
The flow of messages loading and streaming is now more solid.
Everything goes out from `AgentChatAiSdkStreamEffect`, whether loaded
from the DB or streaming directly, and every consumers is using only one
atom `agentChatMessagesComponentFamilyState`
## Data sync effect with callbacks new hook
See
`packages/twenty-front/src/modules/apollo/hooks/useQueryWithCallbacks.ts`
which allows to fix Apollo v4 migration leftovers and is an
implementation of the pattern we talked about with @charlesBochet
We could refine this pattern in another PR.
# Before
https://github.com/user-attachments/assets/84e7a96f-6790-405d-8a73-2dacbf783be5
# After
https://github.com/user-attachments/assets/4c692e3a-2413-4513-abcc-44d0da311203
Co-authored-by: Charles Bochet <charles@twenty.com>
