Skip to content

refactor: use savedMessageId directly instead of re-fetching messages #803

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
larryro merged 1 commit into from
Mar 17, 2026
Merged

refactor: use savedMessageId directly instead of re-fetching messages #803

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
2 changes: 1 addition & 1 deletion services/platform/convex/agent_tools/workflows/internal_mutations.ts
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -136,7 +136,7 @@ export const triggerWorkflowCompletionResponse = internalMutation({
promptMessageId,
maxSteps: 20,
userId: thread?.userId,
deadlineMs: Date.now() + 60_000,
deadlineMs: Date.now() + 420_000,
},
);
},
Expand Down
107 changes: 40 additions & 67 deletions services/platform/convex/lib/agent_completion/on_agent_complete.ts
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -7,11 +7,9 @@
* This function runs in action context and calls mutations as needed.
*/

import { listMessages } from '@convex-dev/agent';

import type { ActionCtx } from '../../_generated/server';

import { components, internal } from '../../_generated/api';
import { internal } from '../../_generated/api';
import { createDebugLog } from '../debug_log';

const debugLog = createDebugLog('DEBUG_AGENT_COMPLETION', '[AgentCompletion]');
Expand All @@ -26,6 +24,7 @@ type Usage = {

export interface AgentResponseResult {
threadId: string;
messageId?: string;
text?: string;
model?: string;
provider?: string;
Expand Down Expand Up @@ -78,74 +77,48 @@ export async function onAgentComplete(
});

// Step 1: Save message metadata (unless skipped)
if (!options?.skipMetadata && result.usage) {
try {
// Find the first assistant message in the current response
// This matches the UIMessage.id logic used by @convex-dev/agent
const messages = await listMessages(ctx, components.agent, {
threadId,
paginationOpts: { cursor: null, numItems: 20 },
excludeToolMessages: false,
});

const sortedMessages = messages.page.sort(
(a, b) => b._creationTime - a._creationTime,
);

const latestAssistantMessage = sortedMessages.find(
(msg) => msg.message?.role === 'assistant',
);

if (latestAssistantMessage) {
const currentOrder = latestAssistantMessage.order;

// Find the FIRST message (by stepOrder) in this response group
const messagesInCurrentResponse = sortedMessages
.filter(
(msg) =>
msg.order === currentOrder &&
(msg.message?.role === 'assistant' ||
msg.message?.role === 'tool'),
)
.sort((a, b) => a.stepOrder - b.stepOrder);

const firstMessageInResponse = messagesInCurrentResponse[0];

if (firstMessageInResponse) {
await ctx.runMutation(
internal.message_metadata.internal_mutations.saveMessageMetadata,
{
messageId: firstMessageInResponse._id,
threadId,
model: result.model,
provider: result.provider,
inputTokens: result.usage.inputTokens,
outputTokens: result.usage.outputTokens,
totalTokens: result.usage.totalTokens,
reasoningTokens: result.usage.reasoningTokens,
cachedInputTokens: result.usage.cachedInputTokens,
reasoning: result.reasoning,
durationMs: result.durationMs,
timeToFirstTokenMs: result.timeToFirstTokenMs,
toolsUsage: result.toolsUsage,
contextWindow: result.contextWindow,
contextStats: result.contextStats,
},
);

debugLog('Metadata saved', {
if (!options?.skipMetadata) {
const messageId = result.messageId;

if (messageId) {
try {
await ctx.runMutation(
internal.message_metadata.internal_mutations.saveMessageMetadata,
{
messageId,
threadId,
agentType,
messageId: firstMessageInResponse._id,
model: result.model,
});
}
provider: result.provider,
inputTokens: result.usage?.inputTokens,
outputTokens: result.usage?.outputTokens,
totalTokens: result.usage?.totalTokens,
reasoningTokens: result.usage?.reasoningTokens,
cachedInputTokens: result.usage?.cachedInputTokens,
reasoning: result.reasoning,
durationMs: result.durationMs,
timeToFirstTokenMs: result.timeToFirstTokenMs,
toolsUsage: result.toolsUsage,
contextWindow: result.contextWindow,
contextStats: result.contextStats,
},
);

debugLog('Metadata saved', {
threadId,
agentType,
messageId,
model: result.model,
});
} catch (error) {
console.error(`[${agentType}] Failed to save message metadata:`, {
threadId,
error,
});
}
} catch (error) {
// Non-fatal: log and continue
console.error(`[${agentType}] Failed to save message metadata:`, {
} else {
debugLog('No messageId provided, skipping metadata save', {
threadId,
error,
agentType,
});
}
}
Expand Down
53 changes: 19 additions & 34 deletions services/platform/convex/lib/agent_response/generate_response.ts
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -460,6 +460,10 @@ export async function generateAgentResponse(
// Track time to first token for streaming
let firstTokenTime: number | null = null;

// The first saved message ID for this generation, used for metadata and approval linking.
// Captured from the agent SDK's savedMessages before any retry logic can overwrite `result`.
let savedMessageId: string | undefined;

// Generate response - streaming or non-streaming
let result: {
text?: string;
Expand Down Expand Up @@ -538,6 +542,8 @@ export async function generateAgentResponse(
},
);

savedMessageId = streamResult.savedMessages?.[0]?._id;

// Wait for stream to complete (with timeout)
const [
streamText,
Expand Down Expand Up @@ -736,6 +742,8 @@ export async function generateAgentResponse(
abortController,
);

savedMessageId = generateResult.savedMessages?.[0]?._id;

debugLog('Generate completed', {
threadId,
elapsedMs: Date.now() - startTime,
Expand Down Expand Up @@ -1134,6 +1142,7 @@ export async function generateAgentResponse(
agentType,
result: {
threadId,
messageId: savedMessageId,
text: responseResult.text,
model: actualModel,
provider,
Expand All @@ -1148,43 +1157,19 @@ export async function generateAgentResponse(
});

// Link approvals to message (only for main agent, not sub-agents)
if (!parentThreadId) {
if (!parentThreadId && savedMessageId) {
try {
const messagesResult = await listMessages(ctx, components.agent, {
threadId,
paginationOpts: { cursor: null, numItems: 50 },
excludeToolMessages: false,
});

const latestAssistantMessage = messagesResult.page.find(
(m: MessageDoc) => m.message?.role === 'assistant',
const linkedCount = await ctx.runMutation(
internal.approvals.internal_mutations.linkApprovalsToMessage,
{
threadId,
messageId: savedMessageId,
},
);

if (latestAssistantMessage) {
const currentOrder = latestAssistantMessage.order;
const messagesInSameOrder = messagesResult.page.filter(
(m: MessageDoc) =>
m.order === currentOrder && m.message?.role !== 'user',
);

messagesInSameOrder.sort(
(a: MessageDoc, b: MessageDoc) => a.stepOrder - b.stepOrder,
);
const firstMessageInOrder =
messagesInSameOrder[0] || latestAssistantMessage;

const linkedCount = await ctx.runMutation(
internal.approvals.internal_mutations.linkApprovalsToMessage,
{
threadId,
messageId: firstMessageInOrder._id,
},
if (linkedCount > 0) {
debugLog(
`Linked ${linkedCount} pending approvals to message ${savedMessageId}`,
);
if (linkedCount > 0) {
debugLog(
`Linked ${linkedCount} pending approvals to message ${firstMessageInOrder._id}`,
);
}
}
} catch (error) {
console.error(
Expand Down
Loading