fix: guard small context windows
This commit is contained in:
@@ -4,6 +4,7 @@
|
|||||||
|
|
||||||
- WhatsApp: refactor vCard parsing helper and improve empty contact card summaries. (#624) — thanks @steipete
|
- WhatsApp: refactor vCard parsing helper and improve empty contact card summaries. (#624) — thanks @steipete
|
||||||
- WhatsApp: include phone numbers when multiple contacts are shared. (#625) — thanks @mahmoudashraf93
|
- WhatsApp: include phone numbers when multiple contacts are shared. (#625) — thanks @mahmoudashraf93
|
||||||
|
- Agents: warn on small context windows (<32k) and block unusable ones (<16k). — thanks @steipete
|
||||||
- Pairing: cap pending DM pairing requests at 3 per provider and avoid pairing replies for outbound DMs. — thanks @steipete
|
- Pairing: cap pending DM pairing requests at 3 per provider and avoid pairing replies for outbound DMs. — thanks @steipete
|
||||||
- macOS: replace relay smoke test with version check in packaging script. (#615) — thanks @YuriNachos
|
- macOS: replace relay smoke test with version check in packaging script. (#615) — thanks @YuriNachos
|
||||||
- macOS: avoid clearing Launch at Login during app initialization. (#607) — thanks @wes-davis
|
- macOS: avoid clearing Launch at Login during app initialization. (#607) — thanks @wes-davis
|
||||||
|
|||||||
135
src/agents/context-window-guard.test.ts
Normal file
135
src/agents/context-window-guard.test.ts
Normal file
@@ -0,0 +1,135 @@
|
|||||||
|
import { describe, expect, it } from "vitest";
|
||||||
|
|
||||||
|
import type { ClawdbotConfig } from "../config/config.js";
|
||||||
|
import {
|
||||||
|
CONTEXT_WINDOW_HARD_MIN_TOKENS,
|
||||||
|
CONTEXT_WINDOW_WARN_BELOW_TOKENS,
|
||||||
|
evaluateContextWindowGuard,
|
||||||
|
resolveContextWindowInfo,
|
||||||
|
} from "./context-window-guard.js";
|
||||||
|
|
||||||
|
describe("context-window-guard", () => {
|
||||||
|
it("blocks below 16k (model metadata)", () => {
|
||||||
|
const info = resolveContextWindowInfo({
|
||||||
|
cfg: undefined,
|
||||||
|
provider: "openrouter",
|
||||||
|
modelId: "tiny",
|
||||||
|
modelContextWindow: 8000,
|
||||||
|
defaultTokens: 200_000,
|
||||||
|
});
|
||||||
|
const guard = evaluateContextWindowGuard({ info });
|
||||||
|
expect(guard.source).toBe("model");
|
||||||
|
expect(guard.tokens).toBe(8000);
|
||||||
|
expect(guard.shouldWarn).toBe(true);
|
||||||
|
expect(guard.shouldBlock).toBe(true);
|
||||||
|
});
|
||||||
|
|
||||||
|
it("warns below 32k but does not block at 16k+", () => {
|
||||||
|
const info = resolveContextWindowInfo({
|
||||||
|
cfg: undefined,
|
||||||
|
provider: "openai",
|
||||||
|
modelId: "small",
|
||||||
|
modelContextWindow: 24_000,
|
||||||
|
defaultTokens: 200_000,
|
||||||
|
});
|
||||||
|
const guard = evaluateContextWindowGuard({ info });
|
||||||
|
expect(guard.tokens).toBe(24_000);
|
||||||
|
expect(guard.shouldWarn).toBe(true);
|
||||||
|
expect(guard.shouldBlock).toBe(false);
|
||||||
|
});
|
||||||
|
|
||||||
|
it("does not warn at 32k+ (model metadata)", () => {
|
||||||
|
const info = resolveContextWindowInfo({
|
||||||
|
cfg: undefined,
|
||||||
|
provider: "openai",
|
||||||
|
modelId: "ok",
|
||||||
|
modelContextWindow: 64_000,
|
||||||
|
defaultTokens: 200_000,
|
||||||
|
});
|
||||||
|
const guard = evaluateContextWindowGuard({ info });
|
||||||
|
expect(guard.shouldWarn).toBe(false);
|
||||||
|
expect(guard.shouldBlock).toBe(false);
|
||||||
|
});
|
||||||
|
|
||||||
|
it("uses models.providers.*.models[].contextWindow when present", () => {
|
||||||
|
const cfg = {
|
||||||
|
models: {
|
||||||
|
providers: {
|
||||||
|
openrouter: {
|
||||||
|
baseUrl: "http://localhost",
|
||||||
|
apiKey: "x",
|
||||||
|
models: [
|
||||||
|
{
|
||||||
|
id: "tiny",
|
||||||
|
name: "tiny",
|
||||||
|
reasoning: false,
|
||||||
|
input: ["text"],
|
||||||
|
cost: { input: 0, output: 0, cacheRead: 0, cacheWrite: 0 },
|
||||||
|
contextWindow: 12_000,
|
||||||
|
maxTokens: 256,
|
||||||
|
},
|
||||||
|
],
|
||||||
|
},
|
||||||
|
},
|
||||||
|
},
|
||||||
|
} satisfies ClawdbotConfig;
|
||||||
|
|
||||||
|
const info = resolveContextWindowInfo({
|
||||||
|
cfg,
|
||||||
|
provider: "openrouter",
|
||||||
|
modelId: "tiny",
|
||||||
|
modelContextWindow: undefined,
|
||||||
|
defaultTokens: 200_000,
|
||||||
|
});
|
||||||
|
const guard = evaluateContextWindowGuard({ info });
|
||||||
|
expect(info.source).toBe("modelsConfig");
|
||||||
|
expect(guard.shouldBlock).toBe(true);
|
||||||
|
});
|
||||||
|
|
||||||
|
it("falls back to agents.defaults.contextTokens", () => {
|
||||||
|
const cfg = {
|
||||||
|
agents: { defaults: { contextTokens: 20_000 } },
|
||||||
|
} satisfies ClawdbotConfig;
|
||||||
|
const info = resolveContextWindowInfo({
|
||||||
|
cfg,
|
||||||
|
provider: "anthropic",
|
||||||
|
modelId: "whatever",
|
||||||
|
modelContextWindow: undefined,
|
||||||
|
defaultTokens: 200_000,
|
||||||
|
});
|
||||||
|
const guard = evaluateContextWindowGuard({ info });
|
||||||
|
expect(info.source).toBe("agentContextTokens");
|
||||||
|
expect(guard.shouldWarn).toBe(true);
|
||||||
|
expect(guard.shouldBlock).toBe(false);
|
||||||
|
});
|
||||||
|
|
||||||
|
it("uses default when nothing else is available", () => {
|
||||||
|
const info = resolveContextWindowInfo({
|
||||||
|
cfg: undefined,
|
||||||
|
provider: "anthropic",
|
||||||
|
modelId: "unknown",
|
||||||
|
modelContextWindow: undefined,
|
||||||
|
defaultTokens: 200_000,
|
||||||
|
});
|
||||||
|
const guard = evaluateContextWindowGuard({ info });
|
||||||
|
expect(info.source).toBe("default");
|
||||||
|
expect(guard.shouldWarn).toBe(false);
|
||||||
|
expect(guard.shouldBlock).toBe(false);
|
||||||
|
});
|
||||||
|
|
||||||
|
it("allows overriding thresholds", () => {
|
||||||
|
const info = { tokens: 10_000, source: "model" as const };
|
||||||
|
const guard = evaluateContextWindowGuard({
|
||||||
|
info,
|
||||||
|
warnBelowTokens: 12_000,
|
||||||
|
hardMinTokens: 9_000,
|
||||||
|
});
|
||||||
|
expect(guard.shouldWarn).toBe(true);
|
||||||
|
expect(guard.shouldBlock).toBe(false);
|
||||||
|
});
|
||||||
|
|
||||||
|
it("exports thresholds as expected", () => {
|
||||||
|
expect(CONTEXT_WINDOW_HARD_MIN_TOKENS).toBe(16_000);
|
||||||
|
expect(CONTEXT_WINDOW_WARN_BELOW_TOKENS).toBe(32_000);
|
||||||
|
});
|
||||||
|
});
|
||||||
84
src/agents/context-window-guard.ts
Normal file
84
src/agents/context-window-guard.ts
Normal file
@@ -0,0 +1,84 @@
|
|||||||
|
import type { ClawdbotConfig } from "../config/config.js";
|
||||||
|
|
||||||
|
export const CONTEXT_WINDOW_HARD_MIN_TOKENS = 16_000;
|
||||||
|
export const CONTEXT_WINDOW_WARN_BELOW_TOKENS = 32_000;
|
||||||
|
|
||||||
|
export type ContextWindowSource =
|
||||||
|
| "model"
|
||||||
|
| "modelsConfig"
|
||||||
|
| "agentContextTokens"
|
||||||
|
| "default";
|
||||||
|
|
||||||
|
export type ContextWindowInfo = {
|
||||||
|
tokens: number;
|
||||||
|
source: ContextWindowSource;
|
||||||
|
};
|
||||||
|
|
||||||
|
function normalizePositiveInt(value: unknown): number | null {
|
||||||
|
if (typeof value !== "number" || !Number.isFinite(value)) return null;
|
||||||
|
const int = Math.floor(value);
|
||||||
|
return int > 0 ? int : null;
|
||||||
|
}
|
||||||
|
|
||||||
|
export function resolveContextWindowInfo(params: {
|
||||||
|
cfg: ClawdbotConfig | undefined;
|
||||||
|
provider: string;
|
||||||
|
modelId: string;
|
||||||
|
modelContextWindow?: number;
|
||||||
|
defaultTokens: number;
|
||||||
|
}): ContextWindowInfo {
|
||||||
|
const fromModel = normalizePositiveInt(params.modelContextWindow);
|
||||||
|
if (fromModel) return { tokens: fromModel, source: "model" };
|
||||||
|
|
||||||
|
const fromModelsConfig = (() => {
|
||||||
|
const providers = params.cfg?.models?.providers as
|
||||||
|
| Record<
|
||||||
|
string,
|
||||||
|
{ models?: Array<{ id?: string; contextWindow?: number }> }
|
||||||
|
>
|
||||||
|
| undefined;
|
||||||
|
const providerEntry = providers?.[params.provider];
|
||||||
|
const models = Array.isArray(providerEntry?.models)
|
||||||
|
? providerEntry.models
|
||||||
|
: [];
|
||||||
|
const match = models.find((m) => m?.id === params.modelId);
|
||||||
|
return normalizePositiveInt(match?.contextWindow);
|
||||||
|
})();
|
||||||
|
if (fromModelsConfig)
|
||||||
|
return { tokens: fromModelsConfig, source: "modelsConfig" };
|
||||||
|
|
||||||
|
const fromAgentConfig = normalizePositiveInt(
|
||||||
|
params.cfg?.agents?.defaults?.contextTokens,
|
||||||
|
);
|
||||||
|
if (fromAgentConfig)
|
||||||
|
return { tokens: fromAgentConfig, source: "agentContextTokens" };
|
||||||
|
|
||||||
|
return { tokens: Math.floor(params.defaultTokens), source: "default" };
|
||||||
|
}
|
||||||
|
|
||||||
|
export type ContextWindowGuardResult = ContextWindowInfo & {
|
||||||
|
shouldWarn: boolean;
|
||||||
|
shouldBlock: boolean;
|
||||||
|
};
|
||||||
|
|
||||||
|
export function evaluateContextWindowGuard(params: {
|
||||||
|
info: ContextWindowInfo;
|
||||||
|
warnBelowTokens?: number;
|
||||||
|
hardMinTokens?: number;
|
||||||
|
}): ContextWindowGuardResult {
|
||||||
|
const warnBelow = Math.max(
|
||||||
|
1,
|
||||||
|
Math.floor(params.warnBelowTokens ?? CONTEXT_WINDOW_WARN_BELOW_TOKENS),
|
||||||
|
);
|
||||||
|
const hardMin = Math.max(
|
||||||
|
1,
|
||||||
|
Math.floor(params.hardMinTokens ?? CONTEXT_WINDOW_HARD_MIN_TOKENS),
|
||||||
|
);
|
||||||
|
const tokens = Math.max(0, Math.floor(params.info.tokens));
|
||||||
|
return {
|
||||||
|
...params.info,
|
||||||
|
tokens,
|
||||||
|
shouldWarn: tokens > 0 && tokens < warnBelow,
|
||||||
|
shouldBlock: tokens > 0 && tokens < hardMin,
|
||||||
|
};
|
||||||
|
}
|
||||||
@@ -42,6 +42,12 @@ import {
|
|||||||
markAuthProfileUsed,
|
markAuthProfileUsed,
|
||||||
} from "./auth-profiles.js";
|
} from "./auth-profiles.js";
|
||||||
import type { BashElevatedDefaults } from "./bash-tools.js";
|
import type { BashElevatedDefaults } from "./bash-tools.js";
|
||||||
|
import {
|
||||||
|
CONTEXT_WINDOW_HARD_MIN_TOKENS,
|
||||||
|
CONTEXT_WINDOW_WARN_BELOW_TOKENS,
|
||||||
|
evaluateContextWindowGuard,
|
||||||
|
resolveContextWindowInfo,
|
||||||
|
} from "./context-window-guard.js";
|
||||||
import {
|
import {
|
||||||
DEFAULT_CONTEXT_TOKENS,
|
DEFAULT_CONTEXT_TOKENS,
|
||||||
DEFAULT_MODEL,
|
DEFAULT_MODEL,
|
||||||
@@ -183,41 +189,13 @@ function resolveContextWindowTokens(params: {
|
|||||||
modelId: string;
|
modelId: string;
|
||||||
model: Model<Api> | undefined;
|
model: Model<Api> | undefined;
|
||||||
}): number {
|
}): number {
|
||||||
const fromModel =
|
return resolveContextWindowInfo({
|
||||||
typeof params.model?.contextWindow === "number" &&
|
cfg: params.cfg,
|
||||||
Number.isFinite(params.model.contextWindow) &&
|
provider: params.provider,
|
||||||
params.model.contextWindow > 0
|
modelId: params.modelId,
|
||||||
? params.model.contextWindow
|
modelContextWindow: params.model?.contextWindow,
|
||||||
: undefined;
|
defaultTokens: DEFAULT_CONTEXT_TOKENS,
|
||||||
if (fromModel) return fromModel;
|
}).tokens;
|
||||||
|
|
||||||
const fromModelsConfig = (() => {
|
|
||||||
const providers = params.cfg?.models?.providers as
|
|
||||||
| Record<
|
|
||||||
string,
|
|
||||||
{ models?: Array<{ id?: string; contextWindow?: number }> }
|
|
||||||
>
|
|
||||||
| undefined;
|
|
||||||
const providerEntry = providers?.[params.provider];
|
|
||||||
const models = Array.isArray(providerEntry?.models)
|
|
||||||
? providerEntry.models
|
|
||||||
: [];
|
|
||||||
const match = models.find((m) => m?.id === params.modelId);
|
|
||||||
return typeof match?.contextWindow === "number" && match.contextWindow > 0
|
|
||||||
? match.contextWindow
|
|
||||||
: undefined;
|
|
||||||
})();
|
|
||||||
if (fromModelsConfig) return fromModelsConfig;
|
|
||||||
|
|
||||||
const fromAgentConfig =
|
|
||||||
typeof params.cfg?.agents?.defaults?.contextTokens === "number" &&
|
|
||||||
Number.isFinite(params.cfg.agents.defaults.contextTokens) &&
|
|
||||||
params.cfg.agents.defaults.contextTokens > 0
|
|
||||||
? Math.floor(params.cfg.agents.defaults.contextTokens)
|
|
||||||
: undefined;
|
|
||||||
if (fromAgentConfig) return fromAgentConfig;
|
|
||||||
|
|
||||||
return DEFAULT_CONTEXT_TOKENS;
|
|
||||||
}
|
}
|
||||||
|
|
||||||
function buildContextPruningExtension(params: {
|
function buildContextPruningExtension(params: {
|
||||||
@@ -1086,6 +1064,33 @@ export async function runEmbeddedPiAgent(params: {
|
|||||||
if (!model) {
|
if (!model) {
|
||||||
throw new Error(error ?? `Unknown model: ${provider}/${modelId}`);
|
throw new Error(error ?? `Unknown model: ${provider}/${modelId}`);
|
||||||
}
|
}
|
||||||
|
|
||||||
|
const ctxInfo = resolveContextWindowInfo({
|
||||||
|
cfg: params.config,
|
||||||
|
provider,
|
||||||
|
modelId,
|
||||||
|
modelContextWindow: model.contextWindow,
|
||||||
|
defaultTokens: DEFAULT_CONTEXT_TOKENS,
|
||||||
|
});
|
||||||
|
const ctxGuard = evaluateContextWindowGuard({
|
||||||
|
info: ctxInfo,
|
||||||
|
warnBelowTokens: CONTEXT_WINDOW_WARN_BELOW_TOKENS,
|
||||||
|
hardMinTokens: CONTEXT_WINDOW_HARD_MIN_TOKENS,
|
||||||
|
});
|
||||||
|
if (ctxGuard.shouldWarn) {
|
||||||
|
log.warn(
|
||||||
|
`low context window: ${provider}/${modelId} ctx=${ctxGuard.tokens} (warn<${CONTEXT_WINDOW_WARN_BELOW_TOKENS}) source=${ctxGuard.source}`,
|
||||||
|
);
|
||||||
|
}
|
||||||
|
if (ctxGuard.shouldBlock) {
|
||||||
|
log.error(
|
||||||
|
`blocked model (context window too small): ${provider}/${modelId} ctx=${ctxGuard.tokens} (min=${CONTEXT_WINDOW_HARD_MIN_TOKENS}) source=${ctxGuard.source}`,
|
||||||
|
);
|
||||||
|
throw new FailoverError(
|
||||||
|
`Model context window too small (${ctxGuard.tokens} tokens). Minimum is ${CONTEXT_WINDOW_HARD_MIN_TOKENS}.`,
|
||||||
|
{ reason: "unknown", provider, model: modelId },
|
||||||
|
);
|
||||||
|
}
|
||||||
const authStore = ensureAuthProfileStore(agentDir);
|
const authStore = ensureAuthProfileStore(agentDir);
|
||||||
const explicitProfileId = params.authProfileId?.trim();
|
const explicitProfileId = params.authProfileId?.trim();
|
||||||
const profileOrder = resolveAuthProfileOrder({
|
const profileOrder = resolveAuthProfileOrder({
|
||||||
|
|||||||
Reference in New Issue
Block a user