@huggingface/tasks 0.20.23 → 0.20.25

package/dist/commonjs/eval.d.ts

@@ -92,5 +92,15 @@ export declare const EVALUATION_FRAMEWORKS: {
         readonly description: "CLAW-Eval is an evaluation framework for assessing LLMs as autonomous agents across 300 human-verified tasks covering communication, finance, and productivity domains.";
         readonly url: "https://github.com/claw-eval/claw-eval";
     };
+    readonly pbench: {
+        readonly name: "pbench";
+        readonly description: "PBench is a multi-level referring expression segmentation benchmark for evaluating vision-language perception across a structured hierarchy of skills.";
+        readonly url: "https://github.com/tiiuae/Falcon-Perception";
+    };
+    readonly wildclawbench: {
+        readonly name: "wildclawbench";
+        readonly description: "WildClawBench is an in-the-wild benchmark for evaluating AI agents in the OpenClaw environment across 60 hand-built, end-to-end tasks spanning productivity, code intelligence, social interaction, search, creative synthesis, and safety domains.";
+        readonly url: "https://github.com/InternLM/WildClawBench";
+    };
 };
 //# sourceMappingURL=eval.d.ts.map

package/dist/commonjs/eval.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"eval.d.ts","sourceRoot":"","sources":["../../src/eval.ts"],"names":[],"mappings":"AAAA;;GAEG;AACH,eAAO,MAAM,qBAAqB;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;CAoGxB,CAAC"}
+{"version":3,"file":"eval.d.ts","sourceRoot":"","sources":["../../src/eval.ts"],"names":[],"mappings":"AAAA;;GAEG;AACH,eAAO,MAAM,qBAAqB;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;CAgHxB,CAAC"}

package/dist/commonjs/eval.js

@@ -95,4 +95,14 @@ exports.EVALUATION_FRAMEWORKS = {
         description: "CLAW-Eval is an evaluation framework for assessing LLMs as autonomous agents across 300 human-verified tasks covering communication, finance, and productivity domains.",
         url: "https://github.com/claw-eval/claw-eval",
     },
+    pbench: {
+        name: "pbench",
+        description: "PBench is a multi-level referring expression segmentation benchmark for evaluating vision-language perception across a structured hierarchy of skills.",
+        url: "https://github.com/tiiuae/Falcon-Perception",
+    },
+    wildclawbench: {
+        name: "wildclawbench",
+        description: "WildClawBench is an in-the-wild benchmark for evaluating AI agents in the OpenClaw environment across 60 hand-built, end-to-end tasks spanning productivity, code intelligence, social interaction, search, creative synthesis, and safety domains.",
+        url: "https://github.com/InternLM/WildClawBench",
+    },
 };

package/dist/commonjs/hardware-amd.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"hardware-amd.d.ts","sourceRoot":"","sources":["../../src/hardware-amd.ts"],"names":[],"mappings":"AAAA,OAAO,KAAK,EAAE,YAAY,EAAE,MAAM,eAAe,CAAC;AAElD,MAAM,WAAW,kBAAmB,SAAQ,YAAY;IACvD;;;;OAIG;IACH,UAAU,EAAE,MAAM,CAAC;CACnB;AAED,eAAO,MAAM,YAAY,EAAE,MAAM,CAAC,MAAM,EAAE,kBAAkB,CAwI3D,CAAC"}
+{"version":3,"file":"hardware-amd.d.ts","sourceRoot":"","sources":["../../src/hardware-amd.ts"],"names":[],"mappings":"AAAA,OAAO,KAAK,EAAE,YAAY,EAAE,MAAM,eAAe,CAAC;AAElD,MAAM,WAAW,kBAAmB,SAAQ,YAAY;IACvD;;;;OAIG;IACH,UAAU,EAAE,MAAM,CAAC;CACnB;AAED,eAAO,MAAM,YAAY,EAAE,MAAM,CAAC,MAAM,EAAE,kBAAkB,CAkJ3D,CAAC"}

package/dist/commonjs/hardware-amd.js

@@ -52,6 +52,16 @@ exports.AMD_GPU_SKUS = {
         memory: [8, 16],
         gfxVersion: "gfx1200",
     },
+    "PRO W7900": {
+        tflops: 122.6,
+        memory: [48],
+        gfxVersion: "gfx1100",
+    },
+    "PRO W7800": {
+        tflops: 90.5,
+        memory: [32, 48],
+        gfxVersion: "gfx1100",
+    },
     "RX 7900 XTX": {
         tflops: 122.8,
         memory: [24],

package/dist/commonjs/local-apps.d.ts

@@ -67,6 +67,7 @@ declare function isLlamaCppGgufModel(model: ModelData): boolean;
 declare function isVllmModel(model: ModelData): boolean;
 declare function isDockerModelRunnerModel(model: ModelData): boolean;
 declare function isUnslothModel(model: ModelData): boolean;
+declare function isToolCallingLocalAgentModel(model: ModelData): boolean;
 /**
  * Add your new local app here.
  *
@@ -241,7 +242,14 @@ export declare const LOCAL_APPS: {
         prettyLabel: string;
         docsUrl: string;
         mainTask: "text-generation";
-        displayOnModelPage: (model: ModelData) => boolean;
+        displayOnModelPage: typeof isToolCallingLocalAgentModel;
+        snippet: (model: ModelData, filepath?: string) => LocalAppSnippet[];
+    };
+    "hermes-agent": {
+        prettyLabel: string;
+        docsUrl: string;
+        mainTask: "text-generation";
+        displayOnModelPage: typeof isToolCallingLocalAgentModel;
         snippet: (model: ModelData, filepath?: string) => LocalAppSnippet[];
     };
 };

package/dist/commonjs/local-apps.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"local-apps.d.ts","sourceRoot":"","sources":["../../src/local-apps.ts"],"names":[],"mappings":"AACA,OAAO,KAAK,EAAE,SAAS,EAAE,MAAM,iBAAiB,CAAC;AACjD,OAAO,KAAK,EAAE,YAAY,EAAE,MAAM,gBAAgB,CAAC;AAKnD,MAAM,WAAW,eAAe;IAC/B;;OAEG;IACH,KAAK,EAAE,MAAM,CAAC;IACd;;OAEG;IACH,KAAK,CAAC,EAAE,MAAM,CAAC;IACf;;OAEG;IACH,OAAO,EAAE,MAAM,GAAG,MAAM,EAAE,CAAC;CAC3B;AAED;;GAEG;AACH,MAAM,MAAM,QAAQ,GAAG;IACtB;;OAEG;IACH,WAAW,EAAE,MAAM,CAAC;IACpB;;OAEG;IACH,OAAO,EAAE,MAAM,CAAC;IAChB;;OAEG;IACH,KAAK,CAAC,EAAE;QAAE,KAAK,EAAE,MAAM,CAAC;QAAC,GAAG,EAAE,MAAM,CAAA;KAAE,EAAE,GAAG,CAAC,CAAC,KAAK,EAAE,SAAS,KAAK;QAAE,KAAK,EAAE,MAAM,CAAC;QAAC,GAAG,EAAE,MAAM,CAAA;KAAE,EAAE,CAAC,CAAC;IACpG;;OAEG;IACH,QAAQ,EAAE,YAAY,CAAC;IACvB;;OAEG;IACH,SAAS,CAAC,EAAE,OAAO,CAAC;IAEpB,UAAU,CAAC,EAAE,OAAO,CAAC;IACrB;;OAEG;IACH,kBAAkB,EAAE,CAAC,KAAK,EAAE,SAAS,KAAK,OAAO,CAAC;CAClD,GAAG,CACD;IACA;;OAEG;IACH,QAAQ,EAAE,CAAC,KAAK,EAAE,SAAS,EAAE,QAAQ,CAAC,EAAE,MAAM,KAAK,GAAG,CAAC;CACtD,GACD;IACA;;;;OAIG;IACH,OAAO,EAAE,CAAC,KAAK,EAAE,SAAS,EAAE,QAAQ,CAAC,EAAE,MAAM,KAAK,MAAM,GAAG,MAAM,EAAE,GAAG,eAAe,GAAG,eAAe,EAAE,CAAC;CACzG,CACH,CAAC;AAsBF,iBAAS,UAAU,CAAC,KAAK,EAAE,SAAS,GAAG,OAAO,CAE7C;AAED,iBAAS,mBAAmB,CAAC,KAAK,EAAE,SAAS,WAE5C;AAED,iBAAS,WAAW,CAAC,KAAK,EAAE,SAAS,GAAG,OAAO,CAU9C;AAED,iBAAS,wBAAwB,CAAC,KAAK,EAAE,SAAS,GAAG,OAAO,CAE3D;AA0BD,iBAAS,cAAc,CAAC,KAAK,EAAE,SAAS,WAEvC;AAiaD;;;;;;;;;;GAUG;AACH,eAAO,MAAM,UAAU;;;;;;yBA/ZS,SAAS,aAAa,MAAM,KAAG,eAAe,EAAE;;;;;;;yBAiDzC,SAAS,aAAa,MAAM,KAAG,eAAe,EAAE;;;;;;;yBA2F3D,SAAS,KAAG,eAAe,EAAE;;;;;;oCA6S3B,SAAS;yBAvPT,SAAS,KAAG,eAAe,EAAE;;;;;;;yBAoF9B,SAAS,KAAG,eAAe,EAAE;;;;;;;yBA7B/B,SAAS,KAAG,eAAe,EAAE;;;;;;;;;;;;;;yBApIzB,SAAS,aAAa,MAAM,KAAG,eAAe,EAAE;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;yBAtDjD,SAAS,aAAa,MAAM,KAAG,MAAM;;;;;;;yBAIpC,SAAS,KAAG,eAAe,EAAE;;;;;;;yBAqSnB,SAAS,aAAa,MAAM,KAAG,MAAM;;;;;;;yBAM9C,SAAS,aAAa,MAAM,KAAG,eAAe,EAAE;;;;;;;yBArDtD,SAAS,aAAa,MAAM,KAAG,eAAe,EAAE;;CA0StC,CAAC;AAErC,MAAM,MAAM,WAAW,GAAG,MAAM,OAAO,UAAU,CAAC"}
+{"version":3,"file":"local-apps.d.ts","sourceRoot":"","sources":["../../src/local-apps.ts"],"names":[],"mappings":"AACA,OAAO,KAAK,EAAE,SAAS,EAAE,MAAM,iBAAiB,CAAC;AACjD,OAAO,KAAK,EAAE,YAAY,EAAE,MAAM,gBAAgB,CAAC;AAKnD,MAAM,WAAW,eAAe;IAC/B;;OAEG;IACH,KAAK,EAAE,MAAM,CAAC;IACd;;OAEG;IACH,KAAK,CAAC,EAAE,MAAM,CAAC;IACf;;OAEG;IACH,OAAO,EAAE,MAAM,GAAG,MAAM,EAAE,CAAC;CAC3B;AAED;;GAEG;AACH,MAAM,MAAM,QAAQ,GAAG;IACtB;;OAEG;IACH,WAAW,EAAE,MAAM,CAAC;IACpB;;OAEG;IACH,OAAO,EAAE,MAAM,CAAC;IAChB;;OAEG;IACH,KAAK,CAAC,EAAE;QAAE,KAAK,EAAE,MAAM,CAAC;QAAC,GAAG,EAAE,MAAM,CAAA;KAAE,EAAE,GAAG,CAAC,CAAC,KAAK,EAAE,SAAS,KAAK;QAAE,KAAK,EAAE,MAAM,CAAC;QAAC,GAAG,EAAE,MAAM,CAAA;KAAE,EAAE,CAAC,CAAC;IACpG;;OAEG;IACH,QAAQ,EAAE,YAAY,CAAC;IACvB;;OAEG;IACH,SAAS,CAAC,EAAE,OAAO,CAAC;IAEpB,UAAU,CAAC,EAAE,OAAO,CAAC;IACrB;;OAEG;IACH,kBAAkB,EAAE,CAAC,KAAK,EAAE,SAAS,KAAK,OAAO,CAAC;CAClD,GAAG,CACD;IACA;;OAEG;IACH,QAAQ,EAAE,CAAC,KAAK,EAAE,SAAS,EAAE,QAAQ,CAAC,EAAE,MAAM,KAAK,GAAG,CAAC;CACtD,GACD;IACA;;;;OAIG;IACH,OAAO,EAAE,CAAC,KAAK,EAAE,SAAS,EAAE,QAAQ,CAAC,EAAE,MAAM,KAAK,MAAM,GAAG,MAAM,EAAE,GAAG,eAAe,GAAG,eAAe,EAAE,CAAC;CACzG,CACH,CAAC;AAsBF,iBAAS,UAAU,CAAC,KAAK,EAAE,SAAS,GAAG,OAAO,CAE7C;AAED,iBAAS,mBAAmB,CAAC,KAAK,EAAE,SAAS,WAE5C;AAED,iBAAS,WAAW,CAAC,KAAK,EAAE,SAAS,GAAG,OAAO,CAU9C;AAED,iBAAS,wBAAwB,CAAC,KAAK,EAAE,SAAS,GAAG,OAAO,CAE3D;AA0BD,iBAAS,cAAc,CAAC,KAAK,EAAE,SAAS,WAEvC;AAED,iBAAS,4BAA4B,CAAC,KAAK,EAAE,SAAS,GAAG,OAAO,CAM/D;AA6bD;;;;;;;;;;GAUG;AACH,eAAO,MAAM,UAAU;;;;;;yBA3bS,SAAS,aAAa,MAAM,KAAG,eAAe,EAAE;;;;;;;yBAiDzC,SAAS,aAAa,MAAM,KAAG,eAAe,EAAE;;;;;;;yBA2F3D,SAAS,KAAG,eAAe,EAAE;;;;;;oCAyU3B,SAAS;yBAnRT,SAAS,KAAG,eAAe,EAAE;;;;;;;yBAoF9B,SAAS,KAAG,eAAe,EAAE;;;;;;;yBA7B/B,SAAS,KAAG,eAAe,EAAE;;;;;;;;;;;;;;yBApIzB,SAAS,aAAa,MAAM,KAAG,eAAe,EAAE;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;yBAtDjD,SAAS,aAAa,MAAM,KAAG,MAAM;;;;;;;yBAIpC,SAAS,KAAG,eAAe,EAAE;;;;;;;yBAiUnB,SAAS,aAAa,MAAM,KAAG,MAAM;;;;;;;yBAM9C,SAAS,aAAa,MAAM,KAAG,eAAe,EAAE;;;;;;;yBAnEtD,SAAS,aAAa,MAAM,KAAG,eAAe,EAAE;;;;;;;yBAkCvC,SAAS,aAAa,MAAM,KAAG,eAAe,EAAE;;CA0R/C,CAAC;AAErC,MAAM,MAAM,WAAW,GAAG,MAAM,OAAO,UAAU,CAAC"}

package/dist/commonjs/local-apps.js

@@ -60,6 +60,11 @@ function getChatTemplate(model) {
 function isUnslothModel(model) {
     return model.tags.includes("unsloth") || isLlamaCppGgufModel(model);
 }
+function isToolCallingLocalAgentModel(model) {
+    return ((isLlamaCppGgufModel(model) || isMlxModel(model)) &&
+        model.tags.includes("conversational") &&
+        !!getChatTemplate(model)?.includes("tools"));
+}
 function getQuantTag(filepath) {
     const defaultTag = ":{{QUANT_TAG}}";
     if (!filepath) {
@@ -357,11 +362,8 @@ const snippetMlxLm = (model) => {
             : []),
     ];
 };
-const snippetPi = (model, filepath) => {
-    const modelName = model.id.split("/").pop() ?? model.id;
-    const isMLX = isMlxModel(model);
-    // Step 1: Server — differs by backend
-    const serverStep = isMLX
+const getLocalServerStep = (model, filepath) => {
+    return isMlxModel(model)
         ? {
             title: "Start the MLX server",
             setup: "# Install MLX LM:\nuv tool install mlx-lm",
@@ -372,14 +374,18 @@ const snippetPi = (model, filepath) => {
             setup: "# Install llama.cpp:\nbrew install llama.cpp",
             content: `# Start a local OpenAI-compatible server:\nllama-server -hf ${model.id}${getQuantTag(filepath)}`,
         };
-    // Step 2: Pi config — port and provider name differ
+};
+const snippetPi = (model, filepath) => {
+    const isMLX = isMlxModel(model);
+    const modelId = isMLX ? model.id : `${model.id}${getQuantTag(filepath)}`;
+    const serverStep = getLocalServerStep(model, filepath);
     const modelsJson = JSON.stringify({
         providers: {
             [isMLX ? "mlx-lm" : "llama-cpp"]: {
                 baseUrl: "http://localhost:8080/v1",
                 api: "openai-completions",
                 apiKey: "none",
-                models: [{ id: isMLX ? model.id : modelName }],
+                models: [{ id: modelId }],
             },
         },
     }, null, 2);
@@ -396,6 +402,31 @@ const snippetPi = (model, filepath) => {
         },
     ];
 };
+const snippetHermesAgent = (model, filepath) => {
+    const modelId = isMlxModel(model) ? model.id : `${model.id}${getQuantTag(filepath)}`;
+    const serverStep = getLocalServerStep(model, filepath);
+    return [
+        serverStep,
+        {
+            title: "Configure Hermes",
+            setup: [
+                "# Install Hermes:",
+                "curl -fsSL https://hermes-agent.nousresearch.com/install.sh | bash",
+                "hermes setup",
+            ].join("\n"),
+            content: [
+                "# Point Hermes at the local server:",
+                "hermes config set model.provider custom",
+                "hermes config set model.base_url http://127.0.0.1:8080/v1",
+                `hermes config set model.default ${modelId}`,
+            ].join("\n"),
+        },
+        {
+            title: "Run Hermes",
+            content: "hermes",
+        },
+    ];
+};
 const snippetDockerModelRunner = (model, filepath) => {
     // Only add quant tag for GGUF models, not safetensors
     const quantTag = isLlamaCppGgufModel(model) ? getQuantTag(filepath) : "";
@@ -636,9 +667,14 @@ exports.LOCAL_APPS = {
         prettyLabel: "Pi",
         docsUrl: "https://github.com/badlogic/pi-mono",
         mainTask: "text-generation",
-        displayOnModelPage: (model) => (isLlamaCppGgufModel(model) || isMlxModel(model)) &&
-            model.tags.includes("conversational") &&
-            !!getChatTemplate(model)?.includes("tools"),
+        displayOnModelPage: isToolCallingLocalAgentModel,
         snippet: snippetPi,
     },
+    "hermes-agent": {
+        prettyLabel: "Hermes Agent",
+        docsUrl: "https://hermes-agent.nousresearch.com/",
+        mainTask: "text-generation",
+        displayOnModelPage: isToolCallingLocalAgentModel,
+        snippet: snippetHermesAgent,
+    },
 };

package/dist/commonjs/local-apps.spec.js

@@ -123,7 +123,7 @@ curl -X POST "http://localhost:8000/v1/chat/completions" \\
         const snippet = snippetFunc(model);
         (0, vitest_1.expect)(snippet[0].content).toContain(`llama-server -hf bartowski/Llama-3.2-3B-Instruct-GGUF:{{QUANT_TAG}}`);
         (0, vitest_1.expect)(snippet[1].setup).toContain("npm install -g @mariozechner/pi-coding-agent");
-        (0, vitest_1.expect)(snippet[1].content).toContain(`"id": "Llama-3.2-3B-Instruct-GGUF"`);
+        (0, vitest_1.expect)(snippet[1].content).toContain(`"id": "bartowski/Llama-3.2-3B-Instruct-GGUF:{{QUANT_TAG}}"`);
         (0, vitest_1.expect)(snippet[2].content).toContain("pi");
     });
     (0, vitest_1.it)("pi - mlx", async () => {
@@ -147,6 +147,40 @@ curl -X POST "http://localhost:8000/v1/chat/completions" \\
         (0, vitest_1.expect)(snippet[1].content).toContain('"id": "mlx-community/Llama-3.2-3B-Instruct-mlx"');
         (0, vitest_1.expect)(snippet[2].content).toContain("pi");
     });
+    (0, vitest_1.it)("hermes-agent", async () => {
+        const { snippet: snippetFunc } = local_apps_js_1.LOCAL_APPS["hermes-agent"];
+        const model = {
+            id: "bartowski/Llama-3.2-3B-Instruct-GGUF",
+            tags: ["conversational"],
+            gguf: { total: 1, context_length: 4096, chat_template: "{% if tools %}" },
+            inference: "",
+        };
+        const snippet = snippetFunc(model);
+        (0, vitest_1.expect)(snippet[0].content).toContain(`llama-server -hf bartowski/Llama-3.2-3B-Instruct-GGUF:{{QUANT_TAG}}`);
+        (0, vitest_1.expect)(snippet[1].content).toContain("hermes config set model.provider custom");
+        (0, vitest_1.expect)(snippet[1].content).toContain("hermes config set model.base_url http://127.0.0.1:8080/v1");
+        (0, vitest_1.expect)(snippet[1].content).toContain("hermes config set model.default bartowski/Llama-3.2-3B-Instruct-GGUF:{{QUANT_TAG}}");
+        (0, vitest_1.expect)(snippet[2].content).toContain("hermes");
+    });
+    (0, vitest_1.it)("hermes-agent - mlx", async () => {
+        const { snippet: snippetFunc } = local_apps_js_1.LOCAL_APPS["hermes-agent"];
+        const model = {
+            id: "mlx-community/Llama-3.2-3B-Instruct-mlx",
+            tags: ["mlx", "conversational"],
+            pipeline_tag: "text-generation",
+            config: {
+                tokenizer_config: {
+                    chat_template: "{% if tools %}...{% endif %}",
+                },
+            },
+            inference: "",
+        };
+        const snippet = snippetFunc(model);
+        (0, vitest_1.expect)(snippet[0].setup).toContain("uv tool install mlx-lm");
+        (0, vitest_1.expect)(snippet[1].content).toContain("hermes config set model.provider custom");
+        (0, vitest_1.expect)(snippet[1].content).toContain("hermes config set model.default mlx-community/Llama-3.2-3B-Instruct-mlx");
+        (0, vitest_1.expect)(snippet[2].content).toContain("hermes");
+    });
     (0, vitest_1.it)("docker model runner", async () => {
         const { snippet: snippetFunc } = local_apps_js_1.LOCAL_APPS["docker-model-runner"];
         const model = {

package/dist/esm/eval.d.ts

@@ -92,5 +92,15 @@ export declare const EVALUATION_FRAMEWORKS: {
         readonly description: "CLAW-Eval is an evaluation framework for assessing LLMs as autonomous agents across 300 human-verified tasks covering communication, finance, and productivity domains.";
         readonly url: "https://github.com/claw-eval/claw-eval";
     };
+    readonly pbench: {
+        readonly name: "pbench";
+        readonly description: "PBench is a multi-level referring expression segmentation benchmark for evaluating vision-language perception across a structured hierarchy of skills.";
+        readonly url: "https://github.com/tiiuae/Falcon-Perception";
+    };
+    readonly wildclawbench: {
+        readonly name: "wildclawbench";
+        readonly description: "WildClawBench is an in-the-wild benchmark for evaluating AI agents in the OpenClaw environment across 60 hand-built, end-to-end tasks spanning productivity, code intelligence, social interaction, search, creative synthesis, and safety domains.";
+        readonly url: "https://github.com/InternLM/WildClawBench";
+    };
 };
 //# sourceMappingURL=eval.d.ts.map

package/dist/esm/eval.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"eval.d.ts","sourceRoot":"","sources":["../../src/eval.ts"],"names":[],"mappings":"AAAA;;GAEG;AACH,eAAO,MAAM,qBAAqB;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;CAoGxB,CAAC"}
+{"version":3,"file":"eval.d.ts","sourceRoot":"","sources":["../../src/eval.ts"],"names":[],"mappings":"AAAA;;GAEG;AACH,eAAO,MAAM,qBAAqB;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;CAgHxB,CAAC"}

package/dist/esm/eval.js

@@ -92,4 +92,14 @@ export const EVALUATION_FRAMEWORKS = {
         description: "CLAW-Eval is an evaluation framework for assessing LLMs as autonomous agents across 300 human-verified tasks covering communication, finance, and productivity domains.",
         url: "https://github.com/claw-eval/claw-eval",
     },
+    pbench: {
+        name: "pbench",
+        description: "PBench is a multi-level referring expression segmentation benchmark for evaluating vision-language perception across a structured hierarchy of skills.",
+        url: "https://github.com/tiiuae/Falcon-Perception",
+    },
+    wildclawbench: {
+        name: "wildclawbench",
+        description: "WildClawBench is an in-the-wild benchmark for evaluating AI agents in the OpenClaw environment across 60 hand-built, end-to-end tasks spanning productivity, code intelligence, social interaction, search, creative synthesis, and safety domains.",
+        url: "https://github.com/InternLM/WildClawBench",
+    },
 };

package/dist/esm/hardware-amd.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"hardware-amd.d.ts","sourceRoot":"","sources":["../../src/hardware-amd.ts"],"names":[],"mappings":"AAAA,OAAO,KAAK,EAAE,YAAY,EAAE,MAAM,eAAe,CAAC;AAElD,MAAM,WAAW,kBAAmB,SAAQ,YAAY;IACvD;;;;OAIG;IACH,UAAU,EAAE,MAAM,CAAC;CACnB;AAED,eAAO,MAAM,YAAY,EAAE,MAAM,CAAC,MAAM,EAAE,kBAAkB,CAwI3D,CAAC"}
+{"version":3,"file":"hardware-amd.d.ts","sourceRoot":"","sources":["../../src/hardware-amd.ts"],"names":[],"mappings":"AAAA,OAAO,KAAK,EAAE,YAAY,EAAE,MAAM,eAAe,CAAC;AAElD,MAAM,WAAW,kBAAmB,SAAQ,YAAY;IACvD;;;;OAIG;IACH,UAAU,EAAE,MAAM,CAAC;CACnB;AAED,eAAO,MAAM,YAAY,EAAE,MAAM,CAAC,MAAM,EAAE,kBAAkB,CAkJ3D,CAAC"}

package/dist/esm/hardware-amd.js

@@ -49,6 +49,16 @@ export const AMD_GPU_SKUS = {
         memory: [8, 16],
         gfxVersion: "gfx1200",
     },
+    "PRO W7900": {
+        tflops: 122.6,
+        memory: [48],
+        gfxVersion: "gfx1100",
+    },
+    "PRO W7800": {
+        tflops: 90.5,
+        memory: [32, 48],
+        gfxVersion: "gfx1100",
+    },
     "RX 7900 XTX": {
         tflops: 122.8,
         memory: [24],

package/dist/esm/local-apps.d.ts

@@ -67,6 +67,7 @@ declare function isLlamaCppGgufModel(model: ModelData): boolean;
 declare function isVllmModel(model: ModelData): boolean;
 declare function isDockerModelRunnerModel(model: ModelData): boolean;
 declare function isUnslothModel(model: ModelData): boolean;
+declare function isToolCallingLocalAgentModel(model: ModelData): boolean;
 /**
  * Add your new local app here.
  *
@@ -241,7 +242,14 @@ export declare const LOCAL_APPS: {
         prettyLabel: string;
         docsUrl: string;
         mainTask: "text-generation";
-        displayOnModelPage: (model: ModelData) => boolean;
+        displayOnModelPage: typeof isToolCallingLocalAgentModel;
+        snippet: (model: ModelData, filepath?: string) => LocalAppSnippet[];
+    };
+    "hermes-agent": {
+        prettyLabel: string;
+        docsUrl: string;
+        mainTask: "text-generation";
+        displayOnModelPage: typeof isToolCallingLocalAgentModel;
         snippet: (model: ModelData, filepath?: string) => LocalAppSnippet[];
     };
 };

package/dist/esm/local-apps.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"local-apps.d.ts","sourceRoot":"","sources":["../../src/local-apps.ts"],"names":[],"mappings":"AACA,OAAO,KAAK,EAAE,SAAS,EAAE,MAAM,iBAAiB,CAAC;AACjD,OAAO,KAAK,EAAE,YAAY,EAAE,MAAM,gBAAgB,CAAC;AAKnD,MAAM,WAAW,eAAe;IAC/B;;OAEG;IACH,KAAK,EAAE,MAAM,CAAC;IACd;;OAEG;IACH,KAAK,CAAC,EAAE,MAAM,CAAC;IACf;;OAEG;IACH,OAAO,EAAE,MAAM,GAAG,MAAM,EAAE,CAAC;CAC3B;AAED;;GAEG;AACH,MAAM,MAAM,QAAQ,GAAG;IACtB;;OAEG;IACH,WAAW,EAAE,MAAM,CAAC;IACpB;;OAEG;IACH,OAAO,EAAE,MAAM,CAAC;IAChB;;OAEG;IACH,KAAK,CAAC,EAAE;QAAE,KAAK,EAAE,MAAM,CAAC;QAAC,GAAG,EAAE,MAAM,CAAA;KAAE,EAAE,GAAG,CAAC,CAAC,KAAK,EAAE,SAAS,KAAK;QAAE,KAAK,EAAE,MAAM,CAAC;QAAC,GAAG,EAAE,MAAM,CAAA;KAAE,EAAE,CAAC,CAAC;IACpG;;OAEG;IACH,QAAQ,EAAE,YAAY,CAAC;IACvB;;OAEG;IACH,SAAS,CAAC,EAAE,OAAO,CAAC;IAEpB,UAAU,CAAC,EAAE,OAAO,CAAC;IACrB;;OAEG;IACH,kBAAkB,EAAE,CAAC,KAAK,EAAE,SAAS,KAAK,OAAO,CAAC;CAClD,GAAG,CACD;IACA;;OAEG;IACH,QAAQ,EAAE,CAAC,KAAK,EAAE,SAAS,EAAE,QAAQ,CAAC,EAAE,MAAM,KAAK,GAAG,CAAC;CACtD,GACD;IACA;;;;OAIG;IACH,OAAO,EAAE,CAAC,KAAK,EAAE,SAAS,EAAE,QAAQ,CAAC,EAAE,MAAM,KAAK,MAAM,GAAG,MAAM,EAAE,GAAG,eAAe,GAAG,eAAe,EAAE,CAAC;CACzG,CACH,CAAC;AAsBF,iBAAS,UAAU,CAAC,KAAK,EAAE,SAAS,GAAG,OAAO,CAE7C;AAED,iBAAS,mBAAmB,CAAC,KAAK,EAAE,SAAS,WAE5C;AAED,iBAAS,WAAW,CAAC,KAAK,EAAE,SAAS,GAAG,OAAO,CAU9C;AAED,iBAAS,wBAAwB,CAAC,KAAK,EAAE,SAAS,GAAG,OAAO,CAE3D;AA0BD,iBAAS,cAAc,CAAC,KAAK,EAAE,SAAS,WAEvC;AAiaD;;;;;;;;;;GAUG;AACH,eAAO,MAAM,UAAU;;;;;;yBA/ZS,SAAS,aAAa,MAAM,KAAG,eAAe,EAAE;;;;;;;yBAiDzC,SAAS,aAAa,MAAM,KAAG,eAAe,EAAE;;;;;;;yBA2F3D,SAAS,KAAG,eAAe,EAAE;;;;;;oCA6S3B,SAAS;yBAvPT,SAAS,KAAG,eAAe,EAAE;;;;;;;yBAoF9B,SAAS,KAAG,eAAe,EAAE;;;;;;;yBA7B/B,SAAS,KAAG,eAAe,EAAE;;;;;;;;;;;;;;yBApIzB,SAAS,aAAa,MAAM,KAAG,eAAe,EAAE;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;yBAtDjD,SAAS,aAAa,MAAM,KAAG,MAAM;;;;;;;yBAIpC,SAAS,KAAG,eAAe,EAAE;;;;;;;yBAqSnB,SAAS,aAAa,MAAM,KAAG,MAAM;;;;;;;yBAM9C,SAAS,aAAa,MAAM,KAAG,eAAe,EAAE;;;;;;;yBArDtD,SAAS,aAAa,MAAM,KAAG,eAAe,EAAE;;CA0StC,CAAC;AAErC,MAAM,MAAM,WAAW,GAAG,MAAM,OAAO,UAAU,CAAC"}
+{"version":3,"file":"local-apps.d.ts","sourceRoot":"","sources":["../../src/local-apps.ts"],"names":[],"mappings":"AACA,OAAO,KAAK,EAAE,SAAS,EAAE,MAAM,iBAAiB,CAAC;AACjD,OAAO,KAAK,EAAE,YAAY,EAAE,MAAM,gBAAgB,CAAC;AAKnD,MAAM,WAAW,eAAe;IAC/B;;OAEG;IACH,KAAK,EAAE,MAAM,CAAC;IACd;;OAEG;IACH,KAAK,CAAC,EAAE,MAAM,CAAC;IACf;;OAEG;IACH,OAAO,EAAE,MAAM,GAAG,MAAM,EAAE,CAAC;CAC3B;AAED;;GAEG;AACH,MAAM,MAAM,QAAQ,GAAG;IACtB;;OAEG;IACH,WAAW,EAAE,MAAM,CAAC;IACpB;;OAEG;IACH,OAAO,EAAE,MAAM,CAAC;IAChB;;OAEG;IACH,KAAK,CAAC,EAAE;QAAE,KAAK,EAAE,MAAM,CAAC;QAAC,GAAG,EAAE,MAAM,CAAA;KAAE,EAAE,GAAG,CAAC,CAAC,KAAK,EAAE,SAAS,KAAK;QAAE,KAAK,EAAE,MAAM,CAAC;QAAC,GAAG,EAAE,MAAM,CAAA;KAAE,EAAE,CAAC,CAAC;IACpG;;OAEG;IACH,QAAQ,EAAE,YAAY,CAAC;IACvB;;OAEG;IACH,SAAS,CAAC,EAAE,OAAO,CAAC;IAEpB,UAAU,CAAC,EAAE,OAAO,CAAC;IACrB;;OAEG;IACH,kBAAkB,EAAE,CAAC,KAAK,EAAE,SAAS,KAAK,OAAO,CAAC;CAClD,GAAG,CACD;IACA;;OAEG;IACH,QAAQ,EAAE,CAAC,KAAK,EAAE,SAAS,EAAE,QAAQ,CAAC,EAAE,MAAM,KAAK,GAAG,CAAC;CACtD,GACD;IACA;;;;OAIG;IACH,OAAO,EAAE,CAAC,KAAK,EAAE,SAAS,EAAE,QAAQ,CAAC,EAAE,MAAM,KAAK,MAAM,GAAG,MAAM,EAAE,GAAG,eAAe,GAAG,eAAe,EAAE,CAAC;CACzG,CACH,CAAC;AAsBF,iBAAS,UAAU,CAAC,KAAK,EAAE,SAAS,GAAG,OAAO,CAE7C;AAED,iBAAS,mBAAmB,CAAC,KAAK,EAAE,SAAS,WAE5C;AAED,iBAAS,WAAW,CAAC,KAAK,EAAE,SAAS,GAAG,OAAO,CAU9C;AAED,iBAAS,wBAAwB,CAAC,KAAK,EAAE,SAAS,GAAG,OAAO,CAE3D;AA0BD,iBAAS,cAAc,CAAC,KAAK,EAAE,SAAS,WAEvC;AAED,iBAAS,4BAA4B,CAAC,KAAK,EAAE,SAAS,GAAG,OAAO,CAM/D;AA6bD;;;;;;;;;;GAUG;AACH,eAAO,MAAM,UAAU;;;;;;yBA3bS,SAAS,aAAa,MAAM,KAAG,eAAe,EAAE;;;;;;;yBAiDzC,SAAS,aAAa,MAAM,KAAG,eAAe,EAAE;;;;;;;yBA2F3D,SAAS,KAAG,eAAe,EAAE;;;;;;oCAyU3B,SAAS;yBAnRT,SAAS,KAAG,eAAe,EAAE;;;;;;;yBAoF9B,SAAS,KAAG,eAAe,EAAE;;;;;;;yBA7B/B,SAAS,KAAG,eAAe,EAAE;;;;;;;;;;;;;;yBApIzB,SAAS,aAAa,MAAM,KAAG,eAAe,EAAE;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;yBAtDjD,SAAS,aAAa,MAAM,KAAG,MAAM;;;;;;;yBAIpC,SAAS,KAAG,eAAe,EAAE;;;;;;;yBAiUnB,SAAS,aAAa,MAAM,KAAG,MAAM;;;;;;;yBAM9C,SAAS,aAAa,MAAM,KAAG,eAAe,EAAE;;;;;;;yBAnEtD,SAAS,aAAa,MAAM,KAAG,eAAe,EAAE;;;;;;;yBAkCvC,SAAS,aAAa,MAAM,KAAG,eAAe,EAAE;;CA0R/C,CAAC;AAErC,MAAM,MAAM,WAAW,GAAG,MAAM,OAAO,UAAU,CAAC"}

package/dist/esm/local-apps.js

@@ -57,6 +57,11 @@ function getChatTemplate(model) {
 function isUnslothModel(model) {
     return model.tags.includes("unsloth") || isLlamaCppGgufModel(model);
 }
+function isToolCallingLocalAgentModel(model) {
+    return ((isLlamaCppGgufModel(model) || isMlxModel(model)) &&
+        model.tags.includes("conversational") &&
+        !!getChatTemplate(model)?.includes("tools"));
+}
 function getQuantTag(filepath) {
     const defaultTag = ":{{QUANT_TAG}}";
     if (!filepath) {
@@ -354,11 +359,8 @@ const snippetMlxLm = (model) => {
             : []),
     ];
 };
-const snippetPi = (model, filepath) => {
-    const modelName = model.id.split("/").pop() ?? model.id;
-    const isMLX = isMlxModel(model);
-    // Step 1: Server — differs by backend
-    const serverStep = isMLX
+const getLocalServerStep = (model, filepath) => {
+    return isMlxModel(model)
         ? {
             title: "Start the MLX server",
             setup: "# Install MLX LM:\nuv tool install mlx-lm",
@@ -369,14 +371,18 @@ const snippetPi = (model, filepath) => {
             setup: "# Install llama.cpp:\nbrew install llama.cpp",
             content: `# Start a local OpenAI-compatible server:\nllama-server -hf ${model.id}${getQuantTag(filepath)}`,
         };
-    // Step 2: Pi config — port and provider name differ
+};
+const snippetPi = (model, filepath) => {
+    const isMLX = isMlxModel(model);
+    const modelId = isMLX ? model.id : `${model.id}${getQuantTag(filepath)}`;
+    const serverStep = getLocalServerStep(model, filepath);
     const modelsJson = JSON.stringify({
         providers: {
             [isMLX ? "mlx-lm" : "llama-cpp"]: {
                 baseUrl: "http://localhost:8080/v1",
                 api: "openai-completions",
                 apiKey: "none",
-                models: [{ id: isMLX ? model.id : modelName }],
+                models: [{ id: modelId }],
             },
         },
     }, null, 2);
@@ -393,6 +399,31 @@ const snippetPi = (model, filepath) => {
         },
     ];
 };
+const snippetHermesAgent = (model, filepath) => {
+    const modelId = isMlxModel(model) ? model.id : `${model.id}${getQuantTag(filepath)}`;
+    const serverStep = getLocalServerStep(model, filepath);
+    return [
+        serverStep,
+        {
+            title: "Configure Hermes",
+            setup: [
+                "# Install Hermes:",
+                "curl -fsSL https://hermes-agent.nousresearch.com/install.sh | bash",
+                "hermes setup",
+            ].join("\n"),
+            content: [
+                "# Point Hermes at the local server:",
+                "hermes config set model.provider custom",
+                "hermes config set model.base_url http://127.0.0.1:8080/v1",
+                `hermes config set model.default ${modelId}`,
+            ].join("\n"),
+        },
+        {
+            title: "Run Hermes",
+            content: "hermes",
+        },
+    ];
+};
 const snippetDockerModelRunner = (model, filepath) => {
     // Only add quant tag for GGUF models, not safetensors
     const quantTag = isLlamaCppGgufModel(model) ? getQuantTag(filepath) : "";
@@ -633,9 +664,14 @@ export const LOCAL_APPS = {
         prettyLabel: "Pi",
         docsUrl: "https://github.com/badlogic/pi-mono",
         mainTask: "text-generation",
-        displayOnModelPage: (model) => (isLlamaCppGgufModel(model) || isMlxModel(model)) &&
-            model.tags.includes("conversational") &&
-            !!getChatTemplate(model)?.includes("tools"),
+        displayOnModelPage: isToolCallingLocalAgentModel,
         snippet: snippetPi,
     },
+    "hermes-agent": {
+        prettyLabel: "Hermes Agent",
+        docsUrl: "https://hermes-agent.nousresearch.com/",
+        mainTask: "text-generation",
+        displayOnModelPage: isToolCallingLocalAgentModel,
+        snippet: snippetHermesAgent,
+    },
 };

package/dist/esm/local-apps.spec.js

@@ -121,7 +121,7 @@ curl -X POST "http://localhost:8000/v1/chat/completions" \\
         const snippet = snippetFunc(model);
         expect(snippet[0].content).toContain(`llama-server -hf bartowski/Llama-3.2-3B-Instruct-GGUF:{{QUANT_TAG}}`);
         expect(snippet[1].setup).toContain("npm install -g @mariozechner/pi-coding-agent");
-        expect(snippet[1].content).toContain(`"id": "Llama-3.2-3B-Instruct-GGUF"`);
+        expect(snippet[1].content).toContain(`"id": "bartowski/Llama-3.2-3B-Instruct-GGUF:{{QUANT_TAG}}"`);
         expect(snippet[2].content).toContain("pi");
     });
     it("pi - mlx", async () => {
@@ -145,6 +145,40 @@ curl -X POST "http://localhost:8000/v1/chat/completions" \\
         expect(snippet[1].content).toContain('"id": "mlx-community/Llama-3.2-3B-Instruct-mlx"');
         expect(snippet[2].content).toContain("pi");
     });
+    it("hermes-agent", async () => {
+        const { snippet: snippetFunc } = LOCAL_APPS["hermes-agent"];
+        const model = {
+            id: "bartowski/Llama-3.2-3B-Instruct-GGUF",
+            tags: ["conversational"],
+            gguf: { total: 1, context_length: 4096, chat_template: "{% if tools %}" },
+            inference: "",
+        };
+        const snippet = snippetFunc(model);
+        expect(snippet[0].content).toContain(`llama-server -hf bartowski/Llama-3.2-3B-Instruct-GGUF:{{QUANT_TAG}}`);
+        expect(snippet[1].content).toContain("hermes config set model.provider custom");
+        expect(snippet[1].content).toContain("hermes config set model.base_url http://127.0.0.1:8080/v1");
+        expect(snippet[1].content).toContain("hermes config set model.default bartowski/Llama-3.2-3B-Instruct-GGUF:{{QUANT_TAG}}");
+        expect(snippet[2].content).toContain("hermes");
+    });
+    it("hermes-agent - mlx", async () => {
+        const { snippet: snippetFunc } = LOCAL_APPS["hermes-agent"];
+        const model = {
+            id: "mlx-community/Llama-3.2-3B-Instruct-mlx",
+            tags: ["mlx", "conversational"],
+            pipeline_tag: "text-generation",
+            config: {
+                tokenizer_config: {
+                    chat_template: "{% if tools %}...{% endif %}",
+                },
+            },
+            inference: "",
+        };
+        const snippet = snippetFunc(model);
+        expect(snippet[0].setup).toContain("uv tool install mlx-lm");
+        expect(snippet[1].content).toContain("hermes config set model.provider custom");
+        expect(snippet[1].content).toContain("hermes config set model.default mlx-community/Llama-3.2-3B-Instruct-mlx");
+        expect(snippet[2].content).toContain("hermes");
+    });
     it("docker model runner", async () => {
         const { snippet: snippetFunc } = LOCAL_APPS["docker-model-runner"];
         const model = {

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@huggingface/tasks",
-  "version": "0.20.23",
+  "version": "0.20.25",
   "description": "List of ML tasks for huggingface.co/tasks",
   "keywords": [
     "hub",

package/src/eval.ts

@@ -101,4 +101,16 @@ export const EVALUATION_FRAMEWORKS = {
 			"CLAW-Eval is an evaluation framework for assessing LLMs as autonomous agents across 300 human-verified tasks covering communication, finance, and productivity domains.",
 		url: "https://github.com/claw-eval/claw-eval",
 	},
+	pbench: {
+		name: "pbench",
+		description:
+			"PBench is a multi-level referring expression segmentation benchmark for evaluating vision-language perception across a structured hierarchy of skills.",
+		url: "https://github.com/tiiuae/Falcon-Perception",
+	},
+	wildclawbench: {
+		name: "wildclawbench",
+		description:
+			"WildClawBench is an in-the-wild benchmark for evaluating AI agents in the OpenClaw environment across 60 hand-built, end-to-end tasks spanning productivity, code intelligence, social interaction, search, creative synthesis, and safety domains.",
+		url: "https://github.com/InternLM/WildClawBench",
+	},
 } as const;

package/src/hardware-amd.ts

@@ -60,6 +60,16 @@ export const AMD_GPU_SKUS: Record<string, AmdGpuHardwareSpec> = {
 		memory: [8, 16],
 		gfxVersion: "gfx1200",
 	},
+	"PRO W7900": {
+		tflops: 122.6,
+		memory: [48],
+		gfxVersion: "gfx1100",
+	},
+	"PRO W7800": {
+		tflops: 90.5,
+		memory: [32, 48],
+		gfxVersion: "gfx1100",
+	},
 	"RX 7900 XTX": {
 		tflops: 122.8,
 		memory: [24],

package/src/local-apps.spec.ts

@@ -134,7 +134,7 @@ curl -X POST "http://localhost:8000/v1/chat/completions" \\
 
 		expect(snippet[0].content).toContain(`llama-server -hf bartowski/Llama-3.2-3B-Instruct-GGUF:{{QUANT_TAG}}`);
 		expect(snippet[1].setup).toContain("npm install -g @mariozechner/pi-coding-agent");
-		expect(snippet[1].content).toContain(`"id": "Llama-3.2-3B-Instruct-GGUF"`);
+		expect(snippet[1].content).toContain(`"id": "bartowski/Llama-3.2-3B-Instruct-GGUF:{{QUANT_TAG}}"`);
 		expect(snippet[2].content).toContain("pi");
 	});
 
@@ -161,6 +161,46 @@ curl -X POST "http://localhost:8000/v1/chat/completions" \\
 		expect(snippet[2].content).toContain("pi");
 	});
 
+	it("hermes-agent", async () => {
+		const { snippet: snippetFunc } = LOCAL_APPS["hermes-agent"];
+		const model: ModelData = {
+			id: "bartowski/Llama-3.2-3B-Instruct-GGUF",
+			tags: ["conversational"],
+			gguf: { total: 1, context_length: 4096, chat_template: "{% if tools %}" },
+			inference: "",
+		};
+		const snippet = snippetFunc(model);
+
+		expect(snippet[0].content).toContain(`llama-server -hf bartowski/Llama-3.2-3B-Instruct-GGUF:{{QUANT_TAG}}`);
+		expect(snippet[1].content).toContain("hermes config set model.provider custom");
+		expect(snippet[1].content).toContain("hermes config set model.base_url http://127.0.0.1:8080/v1");
+		expect(snippet[1].content).toContain(
+			"hermes config set model.default bartowski/Llama-3.2-3B-Instruct-GGUF:{{QUANT_TAG}}",
+		);
+		expect(snippet[2].content).toContain("hermes");
+	});
+
+	it("hermes-agent - mlx", async () => {
+		const { snippet: snippetFunc } = LOCAL_APPS["hermes-agent"];
+		const model: ModelData = {
+			id: "mlx-community/Llama-3.2-3B-Instruct-mlx",
+			tags: ["mlx", "conversational"],
+			pipeline_tag: "text-generation",
+			config: {
+				tokenizer_config: {
+					chat_template: "{% if tools %}...{% endif %}",
+				},
+			},
+			inference: "",
+		};
+		const snippet = snippetFunc(model);
+
+		expect(snippet[0].setup).toContain("uv tool install mlx-lm");
+		expect(snippet[1].content).toContain("hermes config set model.provider custom");
+		expect(snippet[1].content).toContain("hermes config set model.default mlx-community/Llama-3.2-3B-Instruct-mlx");
+		expect(snippet[2].content).toContain("hermes");
+	});
+
 	it("docker model runner", async () => {
 		const { snippet: snippetFunc } = LOCAL_APPS["docker-model-runner"];
 		const model: ModelData = {

package/src/local-apps.ts

@@ -139,6 +139,14 @@ function isUnslothModel(model: ModelData) {
 	return model.tags.includes("unsloth") || isLlamaCppGgufModel(model);
 }
 
+function isToolCallingLocalAgentModel(model: ModelData): boolean {
+	return (
+		(isLlamaCppGgufModel(model) || isMlxModel(model)) &&
+		model.tags.includes("conversational") &&
+		!!getChatTemplate(model)?.includes("tools")
+	);
+}
+
 function getQuantTag(filepath?: string): string {
 	const defaultTag = ":{{QUANT_TAG}}";
 
@@ -463,12 +471,8 @@ const snippetMlxLm = (model: ModelData): LocalAppSnippet[] => {
 	];
 };
 
-const snippetPi = (model: ModelData, filepath?: string): LocalAppSnippet[] => {
-	const modelName = model.id.split("/").pop() ?? model.id;
-	const isMLX = isMlxModel(model);
-
-	// Step 1: Server — differs by backend
-	const serverStep: LocalAppSnippet = isMLX
+const getLocalServerStep = (model: ModelData, filepath?: string): LocalAppSnippet => {
+	return isMlxModel(model)
 		? {
 				title: "Start the MLX server",
 				setup: "# Install MLX LM:\nuv tool install mlx-lm",
@@ -479,8 +483,13 @@ const snippetPi = (model: ModelData, filepath?: string): LocalAppSnippet[] => {
 				setup: "# Install llama.cpp:\nbrew install llama.cpp",
 				content: `# Start a local OpenAI-compatible server:\nllama-server -hf ${model.id}${getQuantTag(filepath)}`,
 			};
+};
+
+const snippetPi = (model: ModelData, filepath?: string): LocalAppSnippet[] => {
+	const isMLX = isMlxModel(model);
+	const modelId = isMLX ? model.id : `${model.id}${getQuantTag(filepath)}`;
+	const serverStep = getLocalServerStep(model, filepath);
 
-	// Step 2: Pi config — port and provider name differ
 	const modelsJson = JSON.stringify(
 		{
 			providers: {
@@ -488,7 +497,7 @@ const snippetPi = (model: ModelData, filepath?: string): LocalAppSnippet[] => {
 					baseUrl: "http://localhost:8080/v1",
 					api: "openai-completions",
 					apiKey: "none",
-					models: [{ id: isMLX ? model.id : modelName }],
+					models: [{ id: modelId }],
 				},
 			},
 		},
@@ -510,6 +519,33 @@ const snippetPi = (model: ModelData, filepath?: string): LocalAppSnippet[] => {
 	];
 };
 
+const snippetHermesAgent = (model: ModelData, filepath?: string): LocalAppSnippet[] => {
+	const modelId = isMlxModel(model) ? model.id : `${model.id}${getQuantTag(filepath)}`;
+	const serverStep = getLocalServerStep(model, filepath);
+
+	return [
+		serverStep,
+		{
+			title: "Configure Hermes",
+			setup: [
+				"# Install Hermes:",
+				"curl -fsSL https://hermes-agent.nousresearch.com/install.sh | bash",
+				"hermes setup",
+			].join("\n"),
+			content: [
+				"# Point Hermes at the local server:",
+				"hermes config set model.provider custom",
+				"hermes config set model.base_url http://127.0.0.1:8080/v1",
+				`hermes config set model.default ${modelId}`,
+			].join("\n"),
+		},
+		{
+			title: "Run Hermes",
+			content: "hermes",
+		},
+	];
+};
+
 const snippetDockerModelRunner = (model: ModelData, filepath?: string): string => {
 	// Only add quant tag for GGUF models, not safetensors
 	const quantTag = isLlamaCppGgufModel(model) ? getQuantTag(filepath) : "";
@@ -755,12 +791,16 @@ export const LOCAL_APPS = {
 		prettyLabel: "Pi",
 		docsUrl: "https://github.com/badlogic/pi-mono",
 		mainTask: "text-generation",
-		displayOnModelPage: (model) =>
-			(isLlamaCppGgufModel(model) || isMlxModel(model)) &&
-			model.tags.includes("conversational") &&
-			!!getChatTemplate(model)?.includes("tools"),
+		displayOnModelPage: isToolCallingLocalAgentModel,
 		snippet: snippetPi,
 	},
+	"hermes-agent": {
+		prettyLabel: "Hermes Agent",
+		docsUrl: "https://hermes-agent.nousresearch.com/",
+		mainTask: "text-generation",
+		displayOnModelPage: isToolCallingLocalAgentModel,
+		snippet: snippetHermesAgent,
+	},
 } satisfies Record<string, LocalApp>;
 
 export type LocalAppKey = keyof typeof LOCAL_APPS;