praxis-agent 0.63.1 → 0.64.1

package/README.md

@@ -28,8 +28,9 @@ sessions, configuration, or compatibility directories.
 - macOS or Linux
 - Node.js 24 or newer
 - [`ripgrep`](https://github.com/BurntSushi/ripgrep) (`rg`) for the Grep and Glob tools
-- an API key and model ID for an Anthropic, OpenAI-compatible, or OpenAI
-  Responses provider (the stable setup), or the explicitly enabled experimental
+- an API key for the uncustomized built-in Anthropic provider (which can select
+  its default model), or an API key and model ID for any other stable provider,
+  or the explicitly enabled experimental
   ChatGPT-backed Codex subscription integration
 
 Praxis does not use Claude subscription authentication. Claude-shaped message,
@@ -92,13 +93,18 @@ For Anthropic Messages:
 ```sh
 export PRAXIS_PROVIDER="anthropic"
 export PRAXIS_API_KEY="your-api-key"
+# Optional: omit it, or use the exact value "default", for current Opus [1m]
 export PRAXIS_MODEL="claude-sonnet-4-6"
 
 cd /path/to/project
 praxis
 ```
 
-Anthropic models use a 200,000-token context window by default, including
+For the uncustomized built-in `anthropic` provider, omitting `PRAXIS_MODEL` (or
+setting it to the exact `default` alias) selects the current Opus model with a
+1,000,000-token context window (`claude-opus-5[1m]`). Other built-in providers
+and custom providers still require an explicit model ID. Anthropic models use a
+200,000-token context window by default, including
 unknown model IDs. Add the exact terminal `[1m]` suffix (for example,
 `claude-sonnet-4-6[1m]`) to request a 1,000,000-token context window;
 Praxis keeps that selected model public, removes the suffix on the wire, and
@@ -306,7 +312,9 @@ troubleshooting. Run `praxis --help` for the authoritative command surface.
   route, whether primary or fallback, stays sticky only through that logical
   Turn's tool continuations; incompatible routes fail closed, and the next
   independent Turn starts from primary. Recovery may persist only an optional
-  selected model, never provider route or wire state. Anthropic uses a
+  selected model, never provider route or wire state. Only the uncustomized
+  built-in Anthropic provider may omit a model, defaulting to
+  `claude-opus-5[1m]`; all other providers require an explicit model. Anthropic uses a
   200,000-token default or exact terminal `[1m]` model syntax for 1,000,000
   tokens; `PRAXIS_CONTEXT_WINDOW_TOKENS` overrides the advertised window.
 - **Transactional self-update** — `praxis update` verifies the package before

package/dist/evals/project-eval-comparison.js

@@ -432,7 +432,8 @@ export function compareProjectEvalAggregates(baseline, candidate, baselineName,
             model: right.model,
         },
         comparable_run_count: leftRuns.length,
-        passed: right.pass_rate >= left.pass_rate &&
+        passed: regressions.length === 0 &&
+            right.pass_rate >= left.pass_rate &&
             safetyKnown &&
             (rightSafety ?? 0) >= (leftSafety ?? 0),
         regressions,

package/dist/providers/anthropic-model-alias.d.ts

@@ -1,4 +1,6 @@
-export type AnthropicModelAliasOverrides = Readonly<Partial<Record<'sonnet' | 'opus' | 'haiku', string>>>;
+export type AnthropicModelFamily = 'sonnet' | 'opus' | 'haiku';
+export type AnthropicModelAliasOverrides = Readonly<Partial<Record<AnthropicModelFamily, string>>>;
+export declare function resolveAnthropicModelAliasFamily(model: string): AnthropicModelFamily | undefined;
 export declare function anthropicModelAliasOverridesFromEnvironment(environment: Readonly<Record<string, string | undefined>>): AnthropicModelAliasOverrides;
 export declare function resolveAnthropicModelAlias(model: string, overrides?: AnthropicModelAliasOverrides): string;
 //# sourceMappingURL=anthropic-model-alias.d.ts.map

package/dist/providers/anthropic-model-alias.js

@@ -8,6 +8,13 @@ const ENVIRONMENT_KEYS = {
     opus: 'ANTHROPIC_DEFAULT_OPUS_MODEL',
     haiku: 'ANTHROPIC_DEFAULT_HAIKU_MODEL',
 };
+export function resolveAnthropicModelAliasFamily(model) {
+    if (model === 'best' || model === 'default')
+        return 'opus';
+    if (model === 'sonnet' || model === 'opus' || model === 'haiku')
+        return model;
+    return undefined;
+}
 export function anthropicModelAliasOverridesFromEnvironment(environment) {
     const overrides = {};
     for (const family of ['sonnet', 'opus', 'haiku']) {
@@ -23,18 +30,19 @@ export function anthropicModelAliasOverridesFromEnvironment(environment) {
 }
 export function resolveAnthropicModelAlias(model, overrides = {}) {
     const longContext = model.endsWith('[1m]');
-    const family = model === 'best'
-        ? 'opus'
-        : longContext
-            ? model.slice(0, -'[1m]'.length)
-            : model;
-    if (family !== 'sonnet' && family !== 'opus' && family !== 'haiku') {
-        return model;
+    const forcedLongContext = model === 'default';
+    let family = resolveAnthropicModelAliasFamily(model);
+    if (family === undefined && longContext) {
+        const baseModel = model.slice(0, -'[1m]'.length);
+        if (baseModel !== 'best' && baseModel !== 'default')
+            family = resolveAnthropicModelAliasFamily(baseModel);
     }
+    if (family === undefined)
+        return model;
     if (longContext && family === 'haiku')
         return model;
     const resolved = overrides[family] ?? DEFAULTS[family];
-    if (!longContext || resolved.endsWith('[1m]'))
+    if ((!longContext && !forcedLongContext) || resolved.endsWith('[1m]'))
         return resolved;
     return `${resolved}[1m]`;
 }

package/dist/providers/provider-registry.js

@@ -7,7 +7,7 @@ import { NonStreamingFallbackModelProvider } from './non-streaming-fallback-prov
 import { CodexOAuthCredentialManager, } from './codex-oauth.js';
 import { resolveProviderTarget, } from './provider-settings.js';
 import { resolveAnthropicModelSpec } from './anthropic-model-spec.js';
-import { anthropicModelAliasOverridesFromEnvironment, resolveAnthropicModelAlias, } from './anthropic-model-alias.js';
+import { anthropicModelAliasOverridesFromEnvironment, resolveAnthropicModelAliasFamily, resolveAnthropicModelAlias, } from './anthropic-model-alias.js';
 import { ProviderAuthenticationError, resolveProviderCredential, } from './provider-auth.js';
 import { parseContextEnvironment, parseProviderEnvironment, } from './environment.js';
 import { createAnthropicPromptCachePolicyResolver, } from './anthropic-prompt-cache.js';
@@ -232,8 +232,8 @@ class NativeProviderRegistry {
         if (this.target.providerId !== 'anthropic' ||
             this.target.protocol !== 'anthropic-messages')
             return false;
-        const family = modelId === 'best' ? 'opus' : modelId;
-        if (family !== 'sonnet' && family !== 'opus' && family !== 'haiku')
+        const family = resolveAnthropicModelAliasFamily(modelId);
+        if (family === undefined)
             return false;
         const override = this.options.anthropicModelAliasOverrides?.[family];
         return override !== undefined && override.trim().length > 0;

package/dist/providers/provider-settings.js

@@ -29,6 +29,7 @@ const BUILT_INS = {
     },
     anthropic: {
         protocol: 'anthropic-messages',
+        defaultModel: 'default',
         profiles: {
             default: {
                 baseUrl: 'https://api.anthropic.com/v1',
@@ -315,7 +316,7 @@ export async function resolveProviderTarget(options) {
             : undefined);
     if (!profile)
         throw new ProviderSettingsError('unknown_profile', `Invalid provider settings: unknown profile ${profileId} for provider ${providerId}`);
-    const modelId = selected.model;
+    const modelId = selected.model ?? definition.defaultModel;
     if (!modelId)
         throw new ProviderSettingsError('model_required', 'Invalid provider settings: model is required');
     const override = environment.PRAXIS_BASE_URL;

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "praxis-agent",
-  "version": "0.63.1",
+  "version": "0.64.1",
   "description": "Local-first, single-user general agent for the command line.",
   "license": "MIT",
   "author": "wuqisen",