Spaces:
Running
Running
| { | |
| "schema_version": "2", | |
| "name": "Decision Model Router", | |
| "references": [ | |
| { | |
| "id": "prompt", | |
| "role": "reference", | |
| "label": "Prompt", | |
| "asset_type": "text", | |
| "inputs": [ | |
| { | |
| "id": "in", | |
| "label": "Text", | |
| "type": "text" | |
| } | |
| ], | |
| "outputs": [ | |
| { | |
| "id": "out", | |
| "label": "Text", | |
| "type": "text" | |
| } | |
| ], | |
| "data": { | |
| "out": "Write a Python function that returns the n-th Fibonacci number in O(log n) time and explain why it works." | |
| } | |
| } | |
| ], | |
| "operators": [ | |
| { | |
| "id": "router", | |
| "role": "operator", | |
| "kind": "space", | |
| "space_id": "vllm-sr/d3-decision-model", | |
| "endpoint": "/decide", | |
| "label": "d3 router: difficulty + domain", | |
| "inputs": [ | |
| { | |
| "id": "state", | |
| "label": "Prompt", | |
| "type": "text", | |
| "required": true | |
| }, | |
| { | |
| "id": "questions", | |
| "label": "Router questions", | |
| "type": "text" | |
| } | |
| ], | |
| "outputs": [ | |
| { | |
| "id": "scores", | |
| "label": "Router verdict", | |
| "type": "json", | |
| "output_index": 1 | |
| } | |
| ], | |
| "data": { | |
| "questions": "{\n \"difficulty\": {\n \"type\": \"score\",\n \"instructions\": \"How hard is it for an AI assistant to answer this prompt correctly? Easy: simple facts, greetings, short rewrites. Hard: multi-step reasoning, non-trivial code, proofs, expert analysis.\",\n \"criteria\": [\n \"easy\",\n \"medium\",\n \"hard\"\n ]\n },\n \"domain\": {\n \"type\": \"choice\",\n \"instructions\": \"What kind of task is this?\",\n \"criteria\": {\n \"code\": \"Programming, debugging, software\",\n \"math\": \"Math, logic, quantitative reasoning\",\n \"general\": \"Facts, advice, everyday questions\",\n \"creative\": \"Writing, stories, rewriting\"\n }\n }\n}" | |
| } | |
| }, | |
| { | |
| "id": "small", | |
| "role": "operator", | |
| "kind": "model", | |
| "model_id": "meta-llama/Llama-3.1-8B-Instruct", | |
| "pipeline_tag": "text-generation", | |
| "endpoint": "chat_completion", | |
| "provider": "deepinfra", | |
| "label": "Small LLM: Llama-3.1-8B (DeepInfra, $0.02 in / $0.05 out per 1M tokens)", | |
| "inputs": [ | |
| { | |
| "id": "text", | |
| "label": "Prompt", | |
| "type": "text", | |
| "required": true | |
| }, | |
| { | |
| "id": "max_tokens", | |
| "label": "Max tokens", | |
| "type": "number" | |
| } | |
| ], | |
| "outputs": [ | |
| { | |
| "id": "out_0", | |
| "label": "Answer", | |
| "type": "markdown", | |
| "output_index": 0 | |
| } | |
| ], | |
| "data": { | |
| "max_tokens": 1024 | |
| } | |
| }, | |
| { | |
| "id": "large", | |
| "role": "operator", | |
| "kind": "model", | |
| "model_id": "openai/gpt-oss-120b", | |
| "pipeline_tag": "text-generation", | |
| "endpoint": "chat_completion", | |
| "provider": "groq", | |
| "label": "Large LLM: gpt-oss-120b (Groq, $0.15 in / $0.75 out per 1M tokens)", | |
| "inputs": [ | |
| { | |
| "id": "text", | |
| "label": "Prompt", | |
| "type": "text", | |
| "required": true | |
| }, | |
| { | |
| "id": "max_tokens", | |
| "label": "Max tokens", | |
| "type": "number" | |
| } | |
| ], | |
| "outputs": [ | |
| { | |
| "id": "out_0", | |
| "label": "Answer", | |
| "type": "markdown", | |
| "output_index": 0 | |
| } | |
| ], | |
| "data": { | |
| "max_tokens": 4096 | |
| } | |
| } | |
| ], | |
| "subjects": [ | |
| { | |
| "id": "verdict", | |
| "role": "subject", | |
| "label": "Router verdict (d3): hard \u2192 large, else small", | |
| "asset_type": "json", | |
| "inputs": [ | |
| { | |
| "id": "in", | |
| "label": "Router verdict (d3): hard \u2192 large, else small", | |
| "type": "json" | |
| } | |
| ], | |
| "outputs": [ | |
| { | |
| "id": "out", | |
| "label": "Router verdict (d3): hard \u2192 large, else small", | |
| "type": "json" | |
| } | |
| ], | |
| "data": {} | |
| }, | |
| { | |
| "id": "small_answer", | |
| "role": "subject", | |
| "label": "Small LLM answer (Llama-3.1-8B, ~13x cheaper per output token)", | |
| "asset_type": "markdown", | |
| "inputs": [ | |
| { | |
| "id": "in", | |
| "label": "Small LLM answer (Llama-3.1-8B, ~13x cheaper per output token)", | |
| "type": "markdown" | |
| } | |
| ], | |
| "outputs": [ | |
| { | |
| "id": "out", | |
| "label": "Small LLM answer (Llama-3.1-8B, ~13x cheaper per output token)", | |
| "type": "markdown" | |
| } | |
| ], | |
| "data": {} | |
| }, | |
| { | |
| "id": "large_answer", | |
| "role": "subject", | |
| "label": "Large LLM answer (gpt-oss-120b)", | |
| "asset_type": "markdown", | |
| "inputs": [ | |
| { | |
| "id": "in", | |
| "label": "Large LLM answer (gpt-oss-120b)", | |
| "type": "markdown" | |
| } | |
| ], | |
| "outputs": [ | |
| { | |
| "id": "out", | |
| "label": "Large LLM answer (gpt-oss-120b)", | |
| "type": "markdown" | |
| } | |
| ], | |
| "data": {} | |
| } | |
| ], | |
| "edges": [ | |
| { | |
| "id": "e0", | |
| "from_node_id": "prompt", | |
| "from_port_id": "out", | |
| "to_node_id": "router", | |
| "to_port_id": "state", | |
| "type": "text" | |
| }, | |
| { | |
| "id": "e1", | |
| "from_node_id": "prompt", | |
| "from_port_id": "out", | |
| "to_node_id": "small", | |
| "to_port_id": "text", | |
| "type": "text" | |
| }, | |
| { | |
| "id": "e2", | |
| "from_node_id": "prompt", | |
| "from_port_id": "out", | |
| "to_node_id": "large", | |
| "to_port_id": "text", | |
| "type": "text" | |
| }, | |
| { | |
| "id": "e3", | |
| "from_node_id": "router", | |
| "from_port_id": "scores", | |
| "to_node_id": "verdict", | |
| "to_port_id": "in", | |
| "type": "json" | |
| }, | |
| { | |
| "id": "e4", | |
| "from_node_id": "small", | |
| "from_port_id": "out_0", | |
| "to_node_id": "small_answer", | |
| "to_port_id": "in", | |
| "type": "markdown" | |
| }, | |
| { | |
| "id": "e5", | |
| "from_node_id": "large", | |
| "from_port_id": "out_0", | |
| "to_node_id": "large_answer", | |
| "to_port_id": "in", | |
| "type": "markdown" | |
| } | |
| ] | |
| } |