smashimo/oh-my-pi
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Merge remote-tracking branch 'upstream/main' into feat/secret-friendly-names

Browse Source
This commit is contained in:
Mathews-Tom
2026-07-01 00:34:03 +05:30
parent 2e611bb078 b6c9747d45
commit d5d3f2fcd7
228 changed files with 9670 additions and 1571 deletions
Download Patch File Download Diff File Expand all files Collapse all files
packages/coding-agent/src/cli/bench-cli.ts
+19 -12
View File
@@ -10,9 +10,10 @@ import type {
Model,
ProviderSessionState,
ServiceTier,
ServiceTierByFamily,
SimpleStreamOptions,
} from "@oh-my-pi/pi-ai";
import { streamSimple } from "@oh-my-pi/pi-ai";
import { resolveModelServiceTier, streamSimple } from "@oh-my-pi/pi-ai";
import { buildModelProviderPriorityRank, type CanonicalModelVariant } from "@oh-my-pi/pi-catalog/identity";
import { replaceTabs, truncateToWidth } from "@oh-my-pi/pi-tui";
import { formatDuration, getProjectDir } from "@oh-my-pi/pi-utils";
@@ -25,7 +26,7 @@ import {
getModelMatchPreferences,
resolveCliModel,
} from "../config/model-resolver";
import { resolveServiceTierSetting } from "../config/service-tier";
import { buildServiceTierByFamily, serviceTierForAllFamilies, serviceTierSettingToTier } from "../config/service-tier";
import { Settings } from "../config/settings";
import benchPrompt from "../prompts/bench.md" with { type: "text" };
import { discoverAuthStorage, loadCliExtensionProviders } from "../sdk";
@@ -106,8 +107,8 @@ export interface BenchSummary {
maxTokens: number;
models: BenchModelReport[];
failures: number;
/** Requested service tier passed to every request; absent when none was requested. Scoped tiers (`openai-only`/`claude-only`) may be dropped per-provider downstream. */
serviceTier?: ServiceTier;
/** Requested per-family service tiers, resolved per model before reaching the wire. */
serviceTierByFamily?: ServiceTierByFamily;
}
type BenchStreamSimple = (
@@ -518,12 +519,18 @@ export async function runBenchCommand(command: BenchCommandArgs, deps: BenchDepe
const runtime = await (deps.createRuntime ?? createDefaultRuntime)();
try {
const targets = resolveBenchModels(command.models, runtime.modelRegistry, runtime.settings, writeStderr);
// Explicit `--service-tier` wins; otherwise fall back to the configured
// `serviceTier` setting (`none`/unset omits the wire field). Scope-aware
// gating to the model's provider happens downstream in the provider layer.
const serviceTierValue = command.flags.serviceTier ?? runtime.settings?.get("serviceTier");
const serviceTier = serviceTierValue ? resolveServiceTierSetting(serviceTierValue, undefined) : undefined;
if (!json && serviceTier) writeStdout(`${chalk.dim(`service tier: ${serviceTier}`)}\n`);
// Explicit `--service-tier` (a single value broadcast across families) wins;
// otherwise fall back to the configured per-family `tier.*` settings. Each
// model resolves its own family's tier below before reaching the wire.
const flagTier = command.flags.serviceTier ? serviceTierSettingToTier(command.flags.serviceTier) : undefined;
const serviceTierByFamily = command.flags.serviceTier
? serviceTierForAllFamilies(flagTier)
: buildServiceTierByFamily(
runtime.settings?.get("tier.openai") ?? "none",
runtime.settings?.get("tier.anthropic") ?? "none",
runtime.settings?.get("tier.google") ?? "none",
);
if (!json && flagTier) writeStdout(`${chalk.dim(`service tier: ${flagTier}`)}\n`);
const reports: BenchModelReport[] = [];
for (const { selector, model, thinking } of targets) {
if (!json) {
@@ -564,7 +571,7 @@ export async function runBenchCommand(command: BenchCommandArgs, deps: BenchDepe
maxTokens,
reasoning: toReasoningEffort(thinking),
disableReasoning: shouldDisableReasoning(thinking) ? true : undefined,
serviceTier,
serviceTier: resolveModelServiceTier(serviceTierByFamily, model),
},
streamFn,
now,
@@ -606,7 +613,7 @@ export async function runBenchCommand(command: BenchCommandArgs, deps: BenchDepe
reports.push(buildModelReport(selector, model, thinking, results));
}
const failures = reports.reduce((sum, report) => sum + report.results.filter(result => !result.ok).length, 0);
const summary: BenchSummary = { runs, maxTokens, models: reports, failures, serviceTier };
const summary: BenchSummary = { runs, maxTokens, models: reports, failures, serviceTierByFamily };
if (json) {
writeStdout(`${JSON.stringify(summary, null, 2)}\n`);
} else if (reports.length > 1 || runs > 1) {
packages/coding-agent/src/cli/tiny-models-cli.ts
+18 -4
View File
@@ -28,12 +28,21 @@ interface ProgressReporter {
interface DownloadResult {
model: TinyLocalModelKey;
ok: boolean;
error?: string;
}
function writeLine(text = ""): void {
process.stdout.write(`${text}\n`);
}
function downloadErrorSummary(error: string | undefined): string | undefined {
return error
?.split(/\r?\n/)
.map(line => line.trim())
.find(line => line.length > 0)
?.replace(/^Error:\s*/, "");
}
export function resolveModels(model: string | undefined): TinyLocalModelKey[] {
if (!model) return [DEFAULT_TINY_TITLE_LOCAL_MODEL_KEY];
// `all` is a prefetch convenience: skip models that fail before load (unsupported
@@ -101,10 +110,15 @@ async function downloadOne(modelKey: TinyLocalModelKey, json: boolean | undefine
const label = getTinyLocalModelSpec(modelKey)?.label ?? modelKey;
if (!json && !process.stdout.isTTY) writeLine(`Downloading ${label} (${modelKey})...`);
const progress = makeProgressReporter(modelKey, json);
const ok = await tinyTitleClient.downloadModel(modelKey, { onProgress: progress.onProgress });
progress.finish(ok);
if (!json && !process.stdout.isTTY) writeLine(ok ? `Downloaded ${label}.` : `Failed to download ${label}.`);
return { model: modelKey, ok };
const result = await tinyTitleClient.downloadModel(modelKey, { onProgress: progress.onProgress });
progress.finish(result.ok);
const error = downloadErrorSummary(result.error);
if (!json && !process.stdout.isTTY) {
writeLine(result.ok ? `Downloaded ${label}.` : `Failed to download ${label}${error ? `: ${error}` : ""}.`);
} else if (!json && !result.ok && error) {
writeLine(`${label} failed: ${error}`);
}
return result.error ? { model: modelKey, ok: result.ok, error: result.error } : { model: modelKey, ok: result.ok };
}
export async function runTinyModelsCommand(command: TinyModelsCommandArgs): Promise<void> {