mirror of
https://github.com/openclaw/openclaw.git
synced 2026-08-25 11:55:47 -06:00
dc93ea6d69
* test(qa): isolate memory scenario evidence * test(qa): follow durable progress completion * test(qa): run doctor migration noninteractively * test(qa): split active memory follow-up * test(qa): wait for preview retirement * test(qa): require ordered preview retirement * test(qa): record ordered durable reply evidence * test(qa): preserve durable delivery budgets * test(qa): isolate durable lifecycle by account * test(qa): scope durable reply assertions by account * test(qa): preserve durable conversation identity
164 lines
10 KiB
YAML
164 lines
10 KiB
YAML
title: Memory tools in channel context
|
|
|
|
scenario:
|
|
id: memory-tools-channel-context
|
|
surface: session-memory
|
|
coverage:
|
|
primary:
|
|
- session-memory.group-memory
|
|
- session-memory.memory-tools
|
|
- session-memory.qa-channel-memory
|
|
secondary:
|
|
- channels.room-session-group-messages
|
|
objective: Verify a shared QA-channel message causally recalls a fact from a nested memory file before the agent replies in the same room.
|
|
successCriteria:
|
|
- The initial model request does not contain the memory-only fact and plans memory_search with the expected arguments.
|
|
- The matching memory_search result plans memory_get for the seeded path, whose correlated result supplies the recalled fact.
|
|
- The ordered tool chain produces exactly one visible final reply in the originating QA room.
|
|
docsRefs:
|
|
- docs/concepts/memory.md
|
|
- docs/concepts/memory-search.md
|
|
codeRefs:
|
|
- extensions/memory-core/src/tools.ts
|
|
- extensions/qa-lab/src/suite.ts
|
|
execution:
|
|
kind: flow
|
|
channel: qa-channel
|
|
providerMode: mock-openai
|
|
retryCount: 0
|
|
summary: Prove a QA-channel group message consumes a matched nested-memory result before one in-room reply.
|
|
config:
|
|
requiredProviderMode: mock-openai
|
|
channelId: qa-memory-room
|
|
channelTitle: QA Memory Room
|
|
expectedMemoryPath: memory/qa-shared-channel.md
|
|
memoryFact: "Hidden QA fact: the project codename is ORBIT-9."
|
|
memoryQuery: "hidden project codename"
|
|
expectedCorpus: null
|
|
expectedNeedle: ORBIT-9
|
|
prompt: "@openclaw Memory tools check: what is the hidden project codename stored only in memory? Use memory tools first."
|
|
promptSnippet: "Memory tools check"
|
|
|
|
flow:
|
|
steps:
|
|
- name: uses memory_search before answering in-channel
|
|
actions:
|
|
- assert:
|
|
expr: "env.providerMode === config.requiredProviderMode"
|
|
message:
|
|
expr: "`expected provider mode ${config.requiredProviderMode}, got ${env.providerMode}`"
|
|
- call: reset
|
|
- call: fs.mkdir
|
|
args:
|
|
- expr: "path.dirname(path.join(env.gateway.workspaceDir, config.expectedMemoryPath))"
|
|
- recursive: true
|
|
- call: fs.writeFile
|
|
args:
|
|
- expr: "path.join(env.gateway.workspaceDir, config.expectedMemoryPath)"
|
|
- expr: "`${config.memoryFact}\\n`"
|
|
- utf8
|
|
- call: forceMemoryIndex
|
|
args:
|
|
- env:
|
|
ref: env
|
|
query:
|
|
expr: config.memoryQuery
|
|
expectedNeedle:
|
|
expr: config.expectedNeedle
|
|
- call: waitForGatewayHealthy
|
|
args:
|
|
- ref: env
|
|
- 60000
|
|
- call: waitForQaChannelReady
|
|
args:
|
|
- ref: env
|
|
- 60000
|
|
- set: requestCursorBeforeInbound
|
|
value:
|
|
expr: "(await fetchJson(`${env.mock.baseUrl}/debug/request-cursor`)).cursor"
|
|
- set: outboundStartIndex
|
|
value:
|
|
expr: "state.getSnapshot().messages.length"
|
|
- set: busCursorBeforeInbound
|
|
value:
|
|
expr: "state.getSnapshot().cursor"
|
|
- sendInbound:
|
|
conversation:
|
|
id:
|
|
expr: config.channelId
|
|
kind: channel
|
|
title:
|
|
expr: config.channelTitle
|
|
senderId: alice
|
|
senderName: Alice
|
|
text:
|
|
expr: config.prompt
|
|
- call: waitForCondition
|
|
saveAs: scenarioRequests
|
|
args:
|
|
- lambda:
|
|
async: true
|
|
expr: "(async () => { const requests = (await fetchJson(`${env.mock.baseUrl}/debug/requests?after=${requestCursorBeforeInbound}`)).filter((request) => String(request.allInputText ?? '').includes(config.promptSnippet)); return requests.length >= 3 ? requests : undefined; })()"
|
|
- expr: liveTurnTimeoutMs(env, 30000)
|
|
- 100
|
|
- assert:
|
|
expr: scenarioRequests.length === 3
|
|
message:
|
|
expr: "`expected one memory_search plan, one matched result request, and one finalization request: ${JSON.stringify(scenarioRequests)}`"
|
|
- set: searchPlanRequest
|
|
value:
|
|
expr: scenarioRequests[0]
|
|
- set: searchResultRequest
|
|
value:
|
|
expr: scenarioRequests[1]
|
|
- set: finalRequest
|
|
value:
|
|
expr: scenarioRequests[2]
|
|
- assert:
|
|
expr: "searchPlanRequest.plannedToolName === 'memory_search' && searchPlanRequest.plannedToolArgs?.query === config.memoryQuery && (searchPlanRequest.plannedToolArgs?.corpus ?? null) === config.expectedCorpus && typeof searchPlanRequest.plannedToolCallId === 'string' && searchPlanRequest.plannedToolCallId.length > 0 && !searchPlanRequest.toolOutputCallId && !String(searchPlanRequest.allInputText ?? '').includes(config.expectedNeedle)"
|
|
message:
|
|
expr: "`initial request exposed the hidden fact or missed the expected memory_search plan: ${JSON.stringify(searchPlanRequest)}`"
|
|
- assert:
|
|
expr: "searchResultRequest.cursor > searchPlanRequest.cursor && searchResultRequest.toolOutputCallId === searchPlanRequest.plannedToolCallId && searchResultRequest.toolOutputStructuredError !== true && String(searchResultRequest.toolOutput ?? '').includes(config.expectedMemoryPath) && String(searchResultRequest.toolOutput ?? '').includes(config.expectedNeedle) && searchResultRequest.plannedToolName === 'memory_get' && searchResultRequest.plannedToolArgs?.path === config.expectedMemoryPath && typeof searchResultRequest.plannedToolCallId === 'string' && searchResultRequest.plannedToolCallId.length > 0 && searchResultRequest.plannedToolCallId !== searchPlanRequest.plannedToolCallId"
|
|
message:
|
|
expr: "`following request did not consume memory_search and plan a distinct memory_get call for the expected path: ${JSON.stringify(searchResultRequest)}`"
|
|
- assert:
|
|
expr: "finalRequest.cursor > searchResultRequest.cursor && finalRequest.toolOutputCallId === searchResultRequest.plannedToolCallId && finalRequest.toolOutputStructuredError !== true && String(finalRequest.toolOutput ?? '').includes(config.expectedMemoryPath) && String(finalRequest.toolOutput ?? '').includes(config.expectedNeedle) && !finalRequest.plannedToolName"
|
|
message:
|
|
expr: "`final request did not consume the matching successful memory_get result: ${JSON.stringify(finalRequest)}`"
|
|
- call: waitForCondition
|
|
saveAs: durableChannelLifecycle
|
|
args:
|
|
- lambda:
|
|
expr: |-
|
|
(() => {
|
|
const snapshot = state.getSnapshot();
|
|
const visible = snapshot.messages.slice(outboundStartIndex).filter((message) => message.accountId === transport.accountId && message.direction === "outbound" && !message.deleted && message.conversation.id === config.channelId && message.conversation.kind === "channel" && message.text.includes(config.expectedNeedle));
|
|
if (visible.length !== 1) return undefined;
|
|
const candidate = visible[0];
|
|
const events = snapshot.events.filter((event) => event.cursor > busCursorBeforeInbound && "message" in event && event.message.accountId === transport.accountId && event.message.direction === "outbound" && event.message.conversation.id === config.channelId && event.message.conversation.kind === "channel" && event.message.text.includes(config.expectedNeedle));
|
|
const finalSent = events.find((event) => event.kind === "outbound-message" && event.message.id === candidate.id);
|
|
if (!finalSent) return undefined;
|
|
const previewRetired = events.find((event) => event.kind === "message-deleted" && event.message.id !== candidate.id && event.cursor < finalSent.cursor);
|
|
if (!previewRetired) return undefined;
|
|
const previewSent = events.find((event) => event.kind === "outbound-message" && event.message.id === previewRetired.message.id && event.cursor < previewRetired.cursor);
|
|
return previewSent ? { message: candidate, previewId: previewRetired.message.id, finalId: candidate.id, previewSentCursor: previewSent.cursor, previewDeletedCursor: previewRetired.cursor, finalSentCursor: finalSent.cursor } : undefined;
|
|
})()
|
|
- expr: liveTurnTimeoutMs(env, 30000)
|
|
- set: durableChannelOutbound
|
|
value:
|
|
expr: durableChannelLifecycle.message
|
|
- set: visibleChannelOutbounds
|
|
value:
|
|
expr: "state.getSnapshot().messages.slice(outboundStartIndex).filter((message) => message.accountId === transport.accountId && message.direction === 'outbound' && !message.deleted && message.conversation.id === config.channelId && message.conversation.kind === 'channel')"
|
|
- assert:
|
|
# Tool-backed finals replace their streamed preview; prove sent -> deleted -> final ordering first.
|
|
expr: "visibleChannelOutbounds.length === 1 && visibleChannelOutbounds[0].id === durableChannelOutbound.id"
|
|
message:
|
|
expr: "`expected exactly one visible QA-room reply after the ordered memory result: ${JSON.stringify(visibleChannelOutbounds)}`"
|
|
- assert:
|
|
expr: "(durableChannelOutbound.text.match(new RegExp(config.expectedNeedle, 'g')) ?? []).length === 1"
|
|
message:
|
|
expr: "`final QA-room reply must contain ${config.expectedNeedle} exactly once: ${durableChannelOutbound.text}`"
|
|
detailsExpr: "`${durableChannelOutbound.text}; path=${config.expectedMemoryPath}; matched=${searchResultRequest.toolOutputCallId === searchPlanRequest.plannedToolCallId}; room=${config.channelId}; lifecycle=${JSON.stringify({ previewId: durableChannelLifecycle.previewId, previewSentCursor: durableChannelLifecycle.previewSentCursor, previewDeletedCursor: durableChannelLifecycle.previewDeletedCursor, finalId: durableChannelLifecycle.finalId, finalSentCursor: durableChannelLifecycle.finalSentCursor })}`"
|