Repository navigation
Normalize unified agent session metadata and AIC accounting #65344
There are no files selected for viewing
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
|
|
@@ -91,6 +91,7 @@ describe("Unified conclusion session", () => { | |
| engine_id: "copilot", | ||
| agent_version: "v2.3.4", | ||
| model: "fixture", | ||
| event_name: "workflow_dispatch", | ||
| secret: "omit", | ||
| }); | ||
| const { events } = collectUnifiedSession({ rootDir: root }); | ||
|
|
@@ -101,14 +102,44 @@ describe("Unified conclusion session", () => { | |
| mcpgVersion: "v0.6.7", | ||
| engineId: "copilot", | ||
| agentVersion: "v2.3.4", | ||
| requestedModel: "fixture", | ||
| triggerType: "workflow_dispatch", | ||
| }, | ||
| provenance: { path: "aw_info.json", component: "workflow" }, | ||
| }); | ||
| expect(events.find(event => event.type === "workflow.info").data).not.toHaveProperty("secret"); | ||
| }); | ||
|
|
||
|
Contributor
There was a problem hiding this comment. [/tdd] New test covers the happy paths for 💡 Suggested additional cases// cache-only usage should still compute AIC
write("evals/evals_token_usage.jsonl", [
{ provider: "openai", model: "gpt-4o-mini", cache_read_input_tokens: 500 },
]);
// negative/non-finite tokens should be rejected, not silently coerced
write("threat-detection/detection_usage.jsonl", [
{ provider: "openai", model: "gpt-4o-mini", input_tokens: -5 },
]);Without these, a regression in the @copilot please address this.
Contributor
Author
There was a problem hiding this comment. Added in |
||
| it("resolves detection and eval AIC per observation without adding overlapping checkpoints", () => { | ||
| write("threat-detection/detection_usage.jsonl", [ | ||
| { provider: "openai", model: "gpt-4o-mini", input_tokens: 1000, output_tokens: 100 }, | ||
| { provider: "openai", model: "unknown-model", input_tokens: 10 }, | ||
| { provider: "openai", model: "gpt-4o-mini", ai_credits: 0, input_tokens: 100 }, | ||
| { usage: { provider: "openai", model: "gpt-4o-mini", ai_credits: 0, input_tokens: 100 } }, | ||
| ]); | ||
| write("evals/evals_token_usage.jsonl", [ | ||
| { provider: "openai", model: "gpt-4o-mini", ai_credits_this_response: 0.25, input_tokens: 100 }, | ||
| { provider: "openai", model: "gpt-4o-mini", input_tokens: 100, output_tokens: 0 }, | ||
| ]); | ||
| const { events } = collectUnifiedSession({ rootDir: root }); | ||
| const detection = events.filter(event => event.type === "usage.report" && event.provenance.phase === "detection"); | ||
| const evals = events.filter(event => event.type === "usage.report" && event.provenance.phase === "evals"); | ||
| expect(detection[0].data.aic).toBeGreaterThan(0); | ||
| expect(detection[0].data.usage).toEqual({ inputTokens: 1000, outputTokens: 100 }); | ||
| expect(detection[1].data).not.toHaveProperty("aic"); | ||
| expect(detection[2].data).toMatchObject({ totalAic: 0 }); | ||
| expect(detection[2].data).not.toHaveProperty("aic"); | ||
| expect(detection[3].data).toMatchObject({ provider: "openai", model: "gpt-4o-mini", totalAic: 0, usage: { inputTokens: 100 } }); | ||
| expect(detection[3].data).not.toHaveProperty("aic"); | ||
| expect(evals[0].data).toMatchObject({ aic: 0.25 }); | ||
| expect(evals[1].data.aic).toBeGreaterThan(0); | ||
| expect(evals[1].data.usage.outputTokens).toBe(0); | ||
| for (const event of [...detection, ...evals]) { | ||
| expect(Object.keys(event.data).every(key => !key.includes("_"))).toBe(true); | ||
| expect(Object.keys(event.data.usage ?? {}).every(key => !key.includes("_"))).toBe(true); | ||
| } | ||
| }); | ||
|
|
||
| it("retains native provenance, equal-time source order, and invalid/absent timestamps without mutation", () => { | ||
| const sources = [ | ||
| { | ||
|
|
@@ -184,7 +215,7 @@ describe("Unified conclusion session", () => { | |
| status: "completed", | ||
| sourceType: "turn.completed", | ||
| numTurns: 1, | ||
| usage: { inputTokens: 16847, outputTokens: 167, cacheReadInputTokens: 8576, cacheCreationInputTokens: 0, reasoningOutputTokens: 0 }, | ||
| }); | ||
| expect(agent.at(-1)).not.toHaveProperty("usage"); | ||
| expect(agent.every(event => event.provenance.path === (canonicalPresent ? "agent-session.jsonl" : "agent-stdio.log"))).toBe(true); | ||
|
|
||
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
|
|
@@ -2,17 +2,18 @@ | |
|
|
||
| /** @typedef {import("./types/agent_session").SessionEvent} SessionEvent */ | ||
| /** @typedef {Record<string, string[]>} Fields */ | ||
|
Contributor
There was a problem hiding this comment. 🤖 Smoke test review comment (Run 37138576096): Good use of named imports from
Contributor
Author
There was a problem hiding this comment. No change was needed in |
||
| const { computeInferenceAIC, findModelPricing } = require("./model_costs.cjs"); | ||
|
|
||
| /** @type {Fields} */ | ||
| const USAGE_FIELDS = { | ||
| totalTokens: ["totalTokens", "total_tokens"], | ||
| inputTokens: ["inputTokens", "input_tokens"], | ||
| outputTokens: ["outputTokens", "output_tokens"], | ||
| reasoningOutputTokens: ["reasoningOutputTokens", "reasoning_output_tokens", "reasoning_tokens"], | ||
| cacheReadInputTokens: ["cacheReadInputTokens", "cache_read_input_tokens", "cache_read_tokens"], | ||
| cacheCreationInputTokens: ["cacheCreationInputTokens", "cache_creation_input_tokens", "cache_write_tokens"], | ||
| inputTokensIncludeCache: ["inputTokensIncludeCache", "input_tokens_include_cache"], | ||
| overflowedTokens: ["overflowedTokens", "overflowed_tokens"], | ||
| }; | ||
|
|
||
| /** @type {Fields} */ | ||
|
|
@@ -121,8 +122,8 @@ const EVENT_FIELDS = { | |
| model: ["model"], | ||
| requestId: ["requestId", "request_id"], | ||
| status: ["status"], | ||
| aic: ["aic", "ai_credits_this_response", "aiCreditsThisResponse"], | ||
| totalAic: ["totalAic", "ai_credits_total", "ai_credits", "aiCredits"], | ||
|
Contributor
Author
There was a problem hiding this comment. Fixed in |
||
| premiumRequests: ["premiumRequests", "premium_requests"], | ||
| durationMs: ["durationMs", "duration_ms"], | ||
| }, | ||
|
|
@@ -143,13 +144,13 @@ const EVENT_FIELDS = { | |
| maliciousPatch: ["maliciousPatch", "malicious_patch"], | ||
| }, | ||
| "workflow.info": { | ||
| engineId: ["engineId", "engine_id", "engine"], | ||
| agentVersion: ["agentVersion", "agent_version"], | ||
| cliVersion: ["cliVersion", "cli_version"], | ||
| awfVersion: ["awfVersion", "awf_version"], | ||
| mcpgVersion: ["mcpgVersion", "mcpg_version", "awmg_version"], | ||
| requestedModel: ["requestedModel", "model"], | ||
| triggerType: ["triggerType", "event_name"], | ||
| workflow: ["workflow", "workflow_name"], | ||
| repository: ["repository"], | ||
| runId: ["runId", "run_id"], | ||
|
|
@@ -173,13 +174,35 @@ function selectFields(value, fields) { | |
| return result; | ||
| } | ||
|
|
||
| /** @param {any} source @param {Record<string, any>} data @param {Record<string, any>} usage */ | ||
| function resolveUsageAic(source, data, usage) { | ||
| if (data.aic !== undefined || data.totalAic !== undefined) return; | ||
| const provider = data.provider ?? source.usage?.provider; | ||
| const model = data.model ?? source.usage?.model; | ||
| if (typeof provider !== "string" || typeof model !== "string" || !findModelPricing(provider, model)) return; | ||
| const tokens = [usage.inputTokens, usage.outputTokens, usage.cacheReadInputTokens, usage.cacheCreationInputTokens, usage.reasoningOutputTokens]; | ||
| if (!tokens.some(value => typeof value === "number" && value > 0) || tokens.some(value => value !== undefined && (typeof value !== "number" || !Number.isFinite(value) || value < 0))) return; | ||
| const aic = computeInferenceAIC({ | ||
| provider, | ||
| model, | ||
| inputTokens: usage.inputTokens ?? 0, | ||
| outputTokens: usage.outputTokens ?? 0, | ||
| cacheReadTokens: usage.cacheReadInputTokens ?? 0, | ||
| cacheWriteTokens: usage.cacheCreationInputTokens ?? 0, | ||
| reasoningTokens: usage.reasoningOutputTokens ?? 0, | ||
| ...(typeof usage.inputTokensIncludeCache === "boolean" ? { inputTokensIncludeCache: usage.inputTokensIncludeCache } : {}), | ||
| }); | ||
| if (Number.isFinite(aic)) data.aic = aic; | ||
| } | ||
|
|
||
| /** | ||
| * Compact known payloads without truncating text or changing native parser traces. | ||
| * Unknown extension data remains opaque for forward compatibility. | ||
| * @param {SessionEvent} event | ||
| * @param {string} [phase] | ||
| * @returns {SessionEvent} | ||
| */ | ||
| function normalizeUnifiedSessionEvent(event, phase) { | ||
| /** @type {any} */ | ||
| const source = event.data; | ||
| const known = Object.hasOwn(EVENT_FIELDS, event.type); | ||
|
|
@@ -193,8 +216,19 @@ function normalizeUnifiedSessionEvent(event) { | |
| } | ||
| if (["session.result", "firewall.token_usage", "usage.report"].includes(event.type)) { | ||
| const usage = selectFields(event.type === "session.result" ? source.usage : (source.usage ?? source), USAGE_FIELDS); | ||
| if (Array.isArray(usage.overflowedTokens)) { | ||
| const aliases = Object.entries(USAGE_FIELDS); | ||
| usage.overflowedTokens = usage.overflowedTokens.map(name => aliases.find(([, names]) => names.includes(name))?.[0] ?? name); | ||
| } | ||
| if (Object.keys(usage).length || (source.usage && typeof source.usage === "object")) data.usage = usage; | ||
| else if (source.usage === null) data.usage = null; | ||
| if (event.type === "usage.report" && source.usage && typeof source.usage === "object") { | ||
| const nested = selectFields(source.usage, EVENT_FIELDS["usage.report"]); | ||
| for (const key of ["provider", "model", "aic", "totalAic", "premiumRequests"]) { | ||
| if (data[key] === undefined && nested[key] !== undefined) data[key] = nested[key]; | ||
| } | ||
| } | ||
| if (event.type === "usage.report" && (phase === "detection" || phase === "evals")) resolveUsageAic(source, data, usage); | ||
| } | ||
| if (event.type === "tool.execution_complete" && (source.is_error === true || source.result?.isError === true || source.result?.is_error === true)) data.isError = true; | ||
| if (event.type === "experiment.assignment") { | ||
|
Contributor
There was a problem hiding this comment. [/codebase-design] 💡 SuggestionIf a future phase (e.g. a new eval/detection source, or @copilot please address this.
Contributor
Author
There was a problem hiding this comment. Fixed in |
||
|
|
||
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Oops, something went wrong.
Oops, something went wrong.
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
🤖 Smoke test review comment (Run 37138576096): The
normalizeSessionUsagefunction correctly handles the object spread — consider adding a comment explaining why arrays are explicitly excluded.There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Added a concise explanation in
f5d8d90: usage is a named-field object, so arrays cannot represent its token fields.