Merge pull request #1 from biggraph/darabos-lynxscribe

data/LynxScribe demo

@@ -0,0 +1,949 @@
+{
+  "env": "LynxScribe",
+  "nodes": [
+    {
+      "id": "Input chat 1",
+      "type": "basic",
+      "data": {
+        "title": "Input chat",
+        "params": {
+          "chat": "mi az elet titka"
+        },
+        "display": null,
+        "error": null,
+        "meta": {
+          "name": "Input chat",
+          "params": {
+            "chat": {
+              "name": "chat",
+              "default": null,
+              "type": {
+                "type": "<class 'str'>"
+              }
+            }
+          },
+          "inputs": {},
+          "outputs": {
+            "output": {
+              "name": "output",
+              "type": {
+                "type": "None"
+              },
+              "position": "right"
+            }
+          },
+          "type": "basic",
+          "sub_nodes": null
+        }
+      },
+      "position": {
+        "x": -292.9043729325095,
+        "y": 59.46898452253504
+      },
+      "parentId": null
+    },
+    {
+      "id": "View 1",
+      "type": "table_view",
+      "data": {
+        "title": "View",
+        "params": {},
+        "display": {
+          "dataframes": {
+            "df": {
+              "columns": [
+                "response"
+              ],
+              "data": [
+                [
+                  "Az élet titka sok ember számára különböző lehet, és sok tényezőtől függ, mint például a személyes értékek, tapasztalatok és célok. Néhány általános gondolat az élet titkairól:\n\n- **Kapcsolatok**: A szeretet és az emberi kapcsolatok nagyon fontosak, hiszen ezek adhatják az élet értelmét.\n- **Önmegvalósítás**: Az, hogy megtaláljuk a szenvedélyeinket és céljainkat, segíthet abban, hogy boldogan éljünk.\n- **Folyamatos tanulás**: Az élet folyamatos tanulás, amely segít fejlődni és alkalmazkodni a változásokhoz.\n- **Egészség**: A fizikai és mentális egészség megőrzése alapvető az életminőség szempontjából.\n\nEzek persze csak általános nézőpontok, és mindenki másképp találhatja meg a saját életének a titkát. Te mivel kapcsolatban keresed az élet titkát?\n\nPlease visit <a href='https://www.linkedin.com/in/g%c3%a1bor-benedek-95578717' target='_blank'>https://www.linkedin.com/in/g%c3%a1bor-benedek-95578717</a> for further information."
+                ]
+              ]
+            }
+          }
+        },
+        "error": null,
+        "meta": {
+          "name": "View",
+          "params": {},
+          "inputs": {
+            "input": {
+              "name": "input",
+              "type": {
+                "type": "<class 'inspect._empty'>"
+              },
+              "position": "left"
+            }
+          },
+          "outputs": {},
+          "type": "table_view",
+          "sub_nodes": null
+        }
+      },
+      "position": {
+        "x": 472.96030572661607,
+        "y": 44.15182379992555
+      },
+      "parentId": null
+    },
+    {
+      "id": "LLM 1",
+      "type": "basic",
+      "data": {
+        "title": "LLM",
+        "params": {
+          "name": "openai"
+        },
+        "display": null,
+        "error": null,
+        "meta": {
+          "name": "LLM",
+          "params": {
+            "name": {
+              "name": "name",
+              "default": "openai",
+              "type": {
+                "type": "<class 'str'>"
+              }
+            }
+          },
+          "inputs": {},
+          "outputs": {
+            "output": {
+              "name": "output",
+              "type": {
+                "type": "None"
+              },
+              "position": "top"
+            }
+          },
+          "type": "basic",
+          "sub_nodes": null
+        }
+      },
+      "position": {
+        "x": -312.5774211084781,
+        "y": 1093.4019527511366
+      },
+      "parentId": null
+    },
+    {
+      "id": "Scenario selector 1",
+      "type": "basic",
+      "data": {
+        "title": "Scenario selector",
+        "params": {
+          "scenario_file": "/home/darabos/nvme/lynxscribe/examples/chat_api/scenarios.yaml",
+          "node_types": "intent_cluster"
+        },
+        "display": null,
+        "error": null,
+        "meta": {
+          "name": "Scenario selector",
+          "params": {
+            "scenario_file": {
+              "name": "scenario_file",
+              "default": null,
+              "type": {
+                "type": "<class 'str'>"
+              }
+            },
+            "node_types": {
+              "name": "node_types",
+              "default": "intent_cluster",
+              "type": {
+                "type": "<class 'str'>"
+              }
+            }
+          },
+          "inputs": {},
+          "outputs": {
+            "output": {
+              "name": "output",
+              "type": {
+                "type": "None"
+              },
+              "position": "top"
+            }
+          },
+          "type": "basic",
+          "sub_nodes": null
+        }
+      },
+      "position": {
+        "x": -549.1300345090008,
+        "y": 1086.4852248156676
+      },
+      "parentId": null
+    },
+    {
+      "id": "Chat API 1",
+      "type": "basic",
+      "data": {
+        "title": "Chat API",
+        "params": {
+          "model": "gpt-4o-mini"
+        },
+        "display": null,
+        "error": null,
+        "meta": {
+          "name": "Chat API",
+          "params": {
+            "model": {
+              "name": "model",
+              "default": "gpt-4o-mini",
+              "type": {
+                "type": "<class 'str'>"
+              }
+            }
+          },
+          "inputs": {
+            "chatbot": {
+              "name": "chatbot",
+              "type": {
+                "type": "<class 'inspect._empty'>"
+              },
+              "position": "bottom"
+            },
+            "chat_processor": {
+              "name": "chat_processor",
+              "type": {
+                "type": "<class 'inspect._empty'>"
+              },
+              "position": "bottom"
+            },
+            "knowledge_base": {
+              "name": "knowledge_base",
+              "type": {
+                "type": "<class 'inspect._empty'>"
+              },
+              "position": "bottom"
+            }
+          },
+          "outputs": {
+            "output": {
+              "name": "output",
+              "type": {
+                "type": "None"
+              },
+              "position": "top"
+            }
+          },
+          "type": "basic",
+          "sub_nodes": null
+        }
+      },
+      "position": {
+        "x": -46.94726514341976,
+        "y": 235.19823621492515
+      },
+      "parentId": null
+    },
+    {
+      "id": "Knowledge base 1",
+      "type": "basic",
+      "data": {
+        "title": "Knowledge base",
+        "params": {
+          "nodes_path": "/home/darabos/nvme/lynxscribe/examples/chat_api/data/lynx/nodes.pickle",
+          "edges_path": "/home/darabos/nvme/lynxscribe/examples/chat_api/data/lynx/edges.pickle",
+          "template_cluster_path": "/home/darabos/nvme/lynxscribe/examples/chat_api/data/lynx/tempclusters.pickle"
+        },
+        "display": null,
+        "error": null,
+        "meta": {
+          "name": "Knowledge base",
+          "params": {
+            "nodes_path": {
+              "name": "nodes_path",
+              "default": "nodes.pickle",
+              "type": {
+                "type": "<class 'str'>"
+              }
+            },
+            "edges_path": {
+              "name": "edges_path",
+              "default": "edges.pickle",
+              "type": {
+                "type": "<class 'str'>"
+              }
+            },
+            "template_cluster_path": {
+              "name": "template_cluster_path",
+              "default": "tempclusters.pickle",
+              "type": {
+                "type": "<class 'str'>"
+              }
+            }
+          },
+          "inputs": {},
+          "outputs": {
+            "output": {
+              "name": "output",
+              "type": {
+                "type": "None"
+              },
+              "position": "top"
+            }
+          },
+          "type": "basic",
+          "sub_nodes": null
+        }
+      },
+      "position": {
+        "x": 382.20164582795104,
+        "y": 533.2833307141879
+      },
+      "parentId": null
+    },
+    {
+      "id": "RAG chatbot 1",
+      "type": "basic",
+      "data": {
+        "title": "RAG chatbot",
+        "params": {
+          "negative_answer": "I'm sorry, but the data I've been trained on does not contain any information related to your question.",
+          "limits_by_type": "{\"information\": [2, 3], \"summary\": [2, 3]}",
+          "strict_limits": true,
+          "max_results": 5
+        },
+        "display": null,
+        "error": null,
+        "meta": {
+          "name": "RAG chatbot",
+          "params": {
+            "negative_answer": {
+              "name": "negative_answer",
+              "default": "I'm sorry, but the data I've been trained on does not contain any information related to your question.",
+              "type": {
+                "type": "<class 'str'>"
+              }
+            },
+            "limits_by_type": {
+              "name": "limits_by_type",
+              "default": "{}",
+              "type": {
+                "type": "<class 'str'>"
+              }
+            },
+            "strict_limits": {
+              "name": "strict_limits",
+              "default": true,
+              "type": {
+                "type": "<class 'bool'>"
+              }
+            },
+            "max_results": {
+              "name": "max_results",
+              "default": 5,
+              "type": {
+                "type": "<class 'int'>"
+              }
+            }
+          },
+          "inputs": {
+            "rag_graph": {
+              "name": "rag_graph",
+              "type": {
+                "type": "<class 'inspect._empty'>"
+              },
+              "position": "bottom"
+            },
+            "scenario_selector": {
+              "name": "scenario_selector",
+              "type": {
+                "type": "<class 'inspect._empty'>"
+              },
+              "position": "bottom"
+            },
+            "llm": {
+              "name": "llm",
+              "type": {
+                "type": "<class 'inspect._empty'>"
+              },
+              "position": "bottom"
+            }
+          },
+          "outputs": {
+            "output": {
+              "name": "output",
+              "type": {
+                "type": "None"
+              },
+              "position": "top"
+            }
+          },
+          "type": "basic",
+          "sub_nodes": null
+        },
+        "beingResized": false
+      },
+      "position": {
+        "x": -521.6507639530705,
+        "y": 547.294980747757
+      },
+      "parentId": null,
+      "width": 336,
+      "height": 349
+    },
+    {
+      "id": "RAG graph 1",
+      "type": "basic",
+      "data": {
+        "title": "RAG graph",
+        "params": {},
+        "display": null,
+        "error": null,
+        "meta": {
+          "name": "RAG graph",
+          "params": {},
+          "inputs": {
+            "vector_store": {
+              "name": "vector_store",
+              "type": {
+                "type": "<class 'inspect._empty'>"
+              },
+              "position": "bottom"
+            },
+            "text_embedder": {
+              "name": "text_embedder",
+              "type": {
+                "type": "<class 'inspect._empty'>"
+              },
+              "position": "bottom"
+            }
+          },
+          "outputs": {
+            "output": {
+              "name": "output",
+              "type": {
+                "type": "None"
+              },
+              "position": "top"
+            }
+          },
+          "type": "basic",
+          "sub_nodes": null
+        }
+      },
+      "position": {
+        "x": -817.8208895639339,
+        "y": 1014.836542916127
+      },
+      "parentId": null
+    },
+    {
+      "id": "Vector store 1",
+      "type": "basic",
+      "data": {
+        "title": "Vector store",
+        "params": {
+          "name": "chromadb",
+          "collection_name": "lynx"
+        },
+        "display": null,
+        "error": null,
+        "meta": {
+          "name": "Vector store",
+          "params": {
+            "name": {
+              "name": "name",
+              "default": "chromadb",
+              "type": {
+                "type": "<class 'str'>"
+              }
+            },
+            "collection_name": {
+              "name": "collection_name",
+              "default": "lynx",
+              "type": {
+                "type": "<class 'str'>"
+              }
+            }
+          },
+          "inputs": {},
+          "outputs": {
+            "output": {
+              "name": "output",
+              "type": {
+                "type": "None"
+              },
+              "position": "top"
+            }
+          },
+          "type": "basic",
+          "sub_nodes": null
+        },
+        "beingResized": false
+      },
+      "position": {
+        "x": -1053.794625339574,
+        "y": 1347.7711940497127
+      },
+      "parentId": null,
+      "width": 275,
+      "height": 227
+    },
+    {
+      "id": "Text embedder 2",
+      "type": "basic",
+      "data": {
+        "title": "Text embedder",
+        "params": {
+          "model": "text-embedding-ada-002"
+        },
+        "display": null,
+        "error": null,
+        "meta": {
+          "name": "Text embedder",
+          "params": {
+            "model": {
+              "name": "model",
+              "default": "text-embedding-ada-002",
+              "type": {
+                "type": "<class 'str'>"
+              }
+            }
+          },
+          "inputs": {
+            "llm": {
+              "name": "llm",
+              "type": {
+                "type": "<class 'inspect._empty'>"
+              },
+              "position": "bottom"
+            }
+          },
+          "outputs": {
+            "output": {
+              "name": "output",
+              "type": {
+                "type": "None"
+              },
+              "position": "top"
+            }
+          },
+          "type": "basic",
+          "sub_nodes": null
+        }
+      },
+      "position": {
+        "x": -749.98604638686,
+        "y": 1293.5978526690794
+      },
+      "parentId": null
+    },
+    {
+      "id": "LLM 2",
+      "type": "basic",
+      "data": {
+        "title": "LLM",
+        "params": {
+          "name": "openai"
+        },
+        "display": null,
+        "error": null,
+        "meta": {
+          "name": "LLM",
+          "params": {
+            "name": {
+              "name": "name",
+              "default": "openai",
+              "type": {
+                "type": "<class 'str'>"
+              }
+            }
+          },
+          "inputs": {},
+          "outputs": {
+            "output": {
+              "name": "output",
+              "type": {
+                "type": "None"
+              },
+              "position": "top"
+            }
+          },
+          "type": "basic",
+          "sub_nodes": null
+        }
+      },
+      "position": {
+        "x": -714.2838040349482,
+        "y": 1469.7242636905507
+      },
+      "parentId": null
+    },
+    {
+      "id": "Truncate history 1",
+      "type": "basic",
+      "data": {
+        "title": "Truncate history",
+        "params": {
+          "max_tokens": 10000,
+          "language": "English"
+        },
+        "display": null,
+        "error": null,
+        "meta": {
+          "name": "Truncate history",
+          "params": {
+            "max_tokens": {
+              "name": "max_tokens",
+              "default": 10000,
+              "type": {
+                "type": "<class 'int'>"
+              }
+            },
+            "language": {
+              "name": "language",
+              "default": "English",
+              "type": {
+                "type": "<class 'str'>"
+              }
+            }
+          },
+          "inputs": {},
+          "outputs": {
+            "output": {
+              "name": "output",
+              "type": {
+                "type": "None"
+              },
+              "position": "top"
+            }
+          },
+          "type": "basic",
+          "sub_nodes": null
+        }
+      },
+      "position": {
+        "x": 0.08889822620079713,
+        "y": 1044.7639853229612
+      },
+      "parentId": null
+    },
+    {
+      "id": "Chat processor 1",
+      "type": "basic",
+      "data": {
+        "title": "Chat processor",
+        "params": {},
+        "display": null,
+        "error": null,
+        "meta": {
+          "name": "Chat processor",
+          "params": {},
+          "inputs": {
+            "processor": {
+              "name": "processor",
+              "type": {
+                "type": "<class 'inspect._empty'>"
+              },
+              "position": "bottom"
+            }
+          },
+          "outputs": {
+            "output": {
+              "name": "output",
+              "type": {
+                "type": "None"
+              },
+              "position": "top"
+            }
+          },
+          "type": "basic",
+          "sub_nodes": null
+        }
+      },
+      "position": {
+        "x": 266.23062579739207,
+        "y": 931.8796075565143
+      },
+      "parentId": null
+    },
+    {
+      "id": "Test Chat API 1",
+      "type": "basic",
+      "data": {
+        "title": "Test Chat API",
+        "params": {},
+        "display": null,
+        "error": null,
+        "meta": {
+          "name": "Test Chat API",
+          "params": {},
+          "inputs": {
+            "message": {
+              "name": "message",
+              "type": {
+                "type": "<class 'inspect._empty'>"
+              },
+              "position": "left"
+            },
+            "chat_api": {
+              "name": "chat_api",
+              "type": {
+                "type": "<class 'inspect._empty'>"
+              },
+              "position": "bottom"
+            }
+          },
+          "outputs": {
+            "output": {
+              "name": "output",
+              "type": {
+                "type": "None"
+              },
+              "position": "right"
+            }
+          },
+          "type": "basic",
+          "sub_nodes": null
+        }
+      },
+      "position": {
+        "x": 30.070483767962628,
+        "y": 30.871931330853627
+      },
+      "parentId": null
+    },
+    {
+      "id": "Mask 1",
+      "type": "basic",
+      "data": {
+        "title": "Mask",
+        "params": {
+          "name": "email",
+          "regex": "([a-z0-9!#$%&'*+\\/=?^_`{|.}~-]+@(?:[a-z0-9](?:[a-z0-9-]*[a-z0-9])?\\.)+[a-z0-9](?:[a-z0-9-]*[a-z0-9])?)",
+          "exceptions": "[email protected],[email protected],[email protected],[email protected]",
+          "mask_pattern": "masked_email_address_{}"
+        },
+        "display": null,
+        "error": null,
+        "meta": {
+          "name": "Mask",
+          "params": {
+            "name": {
+              "name": "name",
+              "default": "",
+              "type": {
+                "type": "None"
+              }
+            },
+            "regex": {
+              "name": "regex",
+              "default": "",
+              "type": {
+                "type": "None"
+              }
+            },
+            "exceptions": {
+              "name": "exceptions",
+              "default": "",
+              "type": {
+                "type": "None"
+              }
+            },
+            "mask_pattern": {
+              "name": "mask_pattern",
+              "default": "",
+              "type": {
+                "type": "None"
+              }
+            }
+          },
+          "inputs": {},
+          "outputs": {
+            "output": {
+              "name": "output",
+              "type": {
+                "type": "None"
+              },
+              "position": "top"
+            }
+          },
+          "type": "basic",
+          "sub_nodes": null
+        }
+      },
+      "position": {
+        "x": 233.69759202223884,
+        "y": 1041.6145468043276
+      },
+      "parentId": null
+    },
+    {
+      "id": "Mask 2",
+      "type": "basic",
+      "data": {
+        "title": "Mask",
+        "params": {
+          "name": "credit_card",
+          "regex": "((?:(?:\\\\d{4}[- ]?){3}\\\\d{4}|\\\\d{15,16}))(?![\\\\d])",
+          "exceptions": "",
+          "mask_pattern": "masked_credit_card_number_{}"
+        },
+        "display": null,
+        "error": null,
+        "meta": {
+          "name": "Mask",
+          "params": {
+            "name": {
+              "name": "name",
+              "default": "",
+              "type": {
+                "type": "None"
+              }
+            },
+            "regex": {
+              "name": "regex",
+              "default": "",
+              "type": {
+                "type": "None"
+              }
+            },
+            "exceptions": {
+              "name": "exceptions",
+              "default": "",
+              "type": {
+                "type": "None"
+              }
+            },
+            "mask_pattern": {
+              "name": "mask_pattern",
+              "default": "",
+              "type": {
+                "type": "None"
+              }
+            }
+          },
+          "inputs": {},
+          "outputs": {
+            "output": {
+              "name": "output",
+              "type": {
+                "type": "None"
+              },
+              "position": "top"
+            }
+          },
+          "type": "basic",
+          "sub_nodes": null
+        }
+      },
+      "position": {
+        "x": 513.2761671440603,
+        "y": 1034.8547191984255
+      },
+      "parentId": null
+    }
+  ],
+  "edges": [
+    {
+      "id": "xy-edge__Knowledge base 1output-Chat API 1knowledge_base",
+      "source": "Knowledge base 1",
+      "target": "Chat API 1",
+      "sourceHandle": "output",
+      "targetHandle": "knowledge_base"
+    },
+    {
+      "id": "xy-edge__RAG chatbot 1output-Chat API 1chatbot",
+      "source": "RAG chatbot 1",
+      "target": "Chat API 1",
+      "sourceHandle": "output",
+      "targetHandle": "chatbot"
+    },
+    {
+      "id": "xy-edge__LLM 1output-RAG chatbot 1llm",
+      "source": "LLM 1",
+      "target": "RAG chatbot 1",
+      "sourceHandle": "output",
+      "targetHandle": "llm"
+    },
+    {
+      "id": "xy-edge__Scenario selector 1output-RAG chatbot 1scenario_selector",
+      "source": "Scenario selector 1",
+      "target": "RAG chatbot 1",
+      "sourceHandle": "output",
+      "targetHandle": "scenario_selector"
+    },
+    {
+      "id": "xy-edge__RAG graph 1output-RAG chatbot 1rag_graph",
+      "source": "RAG graph 1",
+      "target": "RAG chatbot 1",
+      "sourceHandle": "output",
+      "targetHandle": "rag_graph"
+    },
+    {
+      "id": "xy-edge__Vector store 1output-RAG graph 1vector_store",
+      "source": "Vector store 1",
+      "target": "RAG graph 1",
+      "sourceHandle": "output",
+      "targetHandle": "vector_store"
+    },
+    {
+      "id": "xy-edge__Text embedder 2output-RAG graph 1text_embedder",
+      "source": "Text embedder 2",
+      "target": "RAG graph 1",
+      "sourceHandle": "output",
+      "targetHandle": "text_embedder"
+    },
+    {
+      "id": "xy-edge__LLM 2output-Text embedder 2llm",
+      "source": "LLM 2",
+      "target": "Text embedder 2",
+      "sourceHandle": "output",
+      "targetHandle": "llm"
+    },
+    {
+      "id": "xy-edge__Truncate history 1output-Chat processor 1processor",
+      "source": "Truncate history 1",
+      "target": "Chat processor 1",
+      "sourceHandle": "output",
+      "targetHandle": "processor"
+    },
+    {
+      "id": "xy-edge__Chat processor 1output-Chat API 1chat_processor",
+      "source": "Chat processor 1",
+      "target": "Chat API 1",
+      "sourceHandle": "output",
+      "targetHandle": "chat_processor"
+    },
+    {
+      "id": "xy-edge__Chat API 1output-Test Chat API 1chat_api",
+      "source": "Chat API 1",
+      "target": "Test Chat API 1",
+      "sourceHandle": "output",
+      "targetHandle": "chat_api"
+    },
+    {
+      "id": "xy-edge__Test Chat API 1output-View 1input",
+      "source": "Test Chat API 1",
+      "target": "View 1",
+      "sourceHandle": "output",
+      "targetHandle": "input"
+    },
+    {
+      "id": "xy-edge__Input chat 1output-Test Chat API 1message",
+      "source": "Input chat 1",
+      "target": "Test Chat API 1",
+      "sourceHandle": "output",
+      "targetHandle": "message"
+    },
+    {
+      "id": "xy-edge__Mask 1output-Chat processor 1processor",
+      "source": "Mask 1",
+      "target": "Chat processor 1",
+      "sourceHandle": "output",
+      "targetHandle": "processor"
+    },
+    {
+      "id": "xy-edge__Mask 2output-Chat processor 1processor",
+      "source": "Mask 2",
+      "target": "Chat processor 1",
+      "sourceHandle": "output",
+      "targetHandle": "processor"
+    }
+  ]
+}

requirements.txt

@@ -2,6 +2,7 @@ fastapi
 matplotlib
 networkx
 numpy
+orjson
 pandas
 scipy
 uvicorn[standard]

server/executors/one_by_one.py

@@ -1,8 +1,8 @@
 from .. import ops
 from .. import workspace
-import fastapi
-import json
+import orjson
 import pandas as pd
+import pydantic
 import traceback
 import inspect
 import typing
@@ -37,7 +37,7 @@ def register(env: str, cache: bool = True):
   ops.EXECUTORS[env] = lambda ws: execute(ws, ops.CATALOGS[env], cache=cache)
 
 def get_stages(ws, catalog):
-  '''Inputs on top are batch inputs. We decompose the graph into a DAG of components along these edges.'''
+  '''Inputs on top/bottom are batch inputs. We decompose the graph into a DAG of components along these edges.'''
   nodes = {n.id: n for n in ws.nodes}
   batch_inputs = {}
   inputs = {}
@@ -46,7 +46,7 @@ def get_stages(ws, catalog):
     node = nodes[edge.target]
     op = catalog[node.data.title]
     i = op.inputs[edge.targetHandle]
-    if i.position == 'top':
+    if i.position in 'top or bottom':
       batch_inputs.setdefault(edge.target, []).append(edge.source)
   stages = []
   for bt, bss in batch_inputs.items():
@@ -63,6 +63,15 @@ def get_stages(ws, catalog):
   stages.append(set(nodes))
   return stages
 
+
+def _default_serializer(obj):
+  if isinstance(obj, pydantic.BaseModel):
+    return obj.dict()
+  return {"__nonserializable__": id(obj)}
+
+def make_cache_key(obj):
+  return orjson.dumps(obj, default=_default_serializer)
+
 EXECUTOR_OUTPUT_CACHE = {}
 
 def execute(ws, catalog, cache=None):
@@ -77,7 +86,7 @@ def execute(ws, catalog, cache=None):
     node.data.error = None
     op = catalog[node.data.title]
     # Start tasks for nodes that have no non-batch inputs.
-    if all([i.position == 'top' for i in op.inputs.values()]):
+    if all([i.position in 'top or bottom' for i in op.inputs.values()]):
       tasks[node.id] = [NO_INPUT]
   batch_inputs = {}
   # Run the rest until we run out of tasks.
@@ -99,12 +108,12 @@ def execute(ws, catalog, cache=None):
       for task in ts:
         try:
           inputs = [
-            batch_inputs[(n, i.name)] if i.position == 'top' else task
+            batch_inputs[(n, i.name)] if i.position in 'top or bottom' else task
             for i in op.inputs.values()]
-          if cache:
-            key = json.dumps(fastapi.encoders.jsonable_encoder((inputs, params)))
+          if cache is not None:
+            key = make_cache_key((inputs, params))
             if key not in cache:
-              cache[key] = op.func(*inputs, **params)
+              cache[key] = op(*inputs, **params)
             result = cache[key]
           else:
             result = op(*inputs, **params)
@@ -126,8 +135,9 @@ def execute(ws, catalog, cache=None):
           t = nodes[edge.target]
           op = catalog[t.data.title]
           i = op.inputs[edge.targetHandle]
-          if i.position == 'top':
+          if i.position in 'top or bottom':
             batch_inputs.setdefault((edge.target, edge.targetHandle), []).extend(results)
           else:
             tasks.setdefault(edge.target, []).extend(results)
     tasks = next_stage
+  return contexts

server/lynxscribe_ops.py

@@ -1,52 +1,200 @@
-'''An example of passive ops. Just using LynxKite to describe the configuration of a complex system.'''
-from .ops import passive_op_registration, Parameter as P, MULTI_INPUT
-
-reg = passive_op_registration('LynxScribe')
-reg('Scrape documents', params=[P.basic('url', '')])
-reg('Conversation logs')
-reg('Extract graph', inputs=['input'])
-reg('Compute embeddings', inputs=['input'], params=[P.options('method', ['OpenAI', 'graph', 'Yi-34b']), P.basic('dimensions', 1234)])
-reg('Vector DB', inputs=[MULTI_INPUT], params=[P.options('backend', ['FAISS', 'ANN', 'HNSW'])])
-reg('Chat UI', outputs=[], inputs=['input'])
-reg('Chat backend')
-reg('WhatsApp')
-reg('PII removal', inputs=['input'])
-reg('Intent classification', inputs=['input'])
-reg('System prompt', params=[P.basic('prompt', 'You are a helpful chatbot.')])
-reg('LLM', inputs=[MULTI_INPUT], params=[P.options('backend', ['GPT-4', 'Yi-34b', 'Claude 3 Opus', 'Google Gemini'])])
-
-# From Marton's mock-up.
-yi = 'Yi-34B (triton)'
-reg('Chat Input', params=[
-  P.options('load_mode', ['augmented']),
-  P.options('model', [yi]),
-  P.options('embedder', ['GritLM-7b (triton)']),
-  ])
-reg('k-NN Intent Classifier', inputs=['qa_embs', 'rag_graph'], params=[
-  P.options('distance', ['cosine', 'euclidean']),
-  P.basic('max_dist', 0.3),
-  P.basic('k', 3),
-  P.options('voting', ['most common', 'weighted']),
-  ])
-reg('Chroma Graph RAG Loader', inputs=[], params=[
-  P.options('location', ['GCP']),
-  P.collapsed('bucket', ''),
-  P.collapsed('folder', ''),
-  P.options('embedder', ['GritLM-7b (triton)']),
-  ])
-reg('Scenario Builder', inputs=['input'], params=[
-  P.collapsed('scenario', ''),
-  ])
-reg('Graph RAG Answer', inputs=['qa_embs', 'intent', 'rag_graph', 'prompt_dict'], params=[
-  P.options('answer_llm', [yi]),
-  P.basic('faq_dist', 0.12),
-  P.basic('max_dist', 0.25),
-  P.basic('ctx_tokens', 2800),
-  P.options('distance', ['cosine', 'euclidean']),
-  P.collapsed('graph_rag_params', ''),
-  ])
-reg('Answer Post Processing', inputs=['qa_embs', 'rag_graph'], params=[
-  P.options('distance', ['cosine', 'euclidean']),
-  P.basic('min_conf', 0.78),
-  ])
-reg('Chat Output', inputs=['input'], outputs=[])
+"""
+LynxScribe configuration and testing in LynxKite.
+"""
+from lynxscribe.core.llm.base import get_llm_engine
+from lynxscribe.core.vector_store.base import get_vector_store
+from lynxscribe.common.config import load_config
+from lynxscribe.components.text_embedder import TextEmbedder
+from lynxscribe.components.rag.rag_graph import RAGGraph
+from lynxscribe.components.rag.knowledge_base_graph import PandasKnowledgeBaseGraph
+from lynxscribe.components.rag.rag_chatbot import Scenario, ScenarioSelector, RAGChatbot
+from lynxscribe.components.chat_processor.base import ChatProcessor
+from lynxscribe.components.chat_processor.processors import MaskTemplate, TruncateHistory
+from lynxscribe.components.chat_api import ChatAPI, ChatAPIRequest, ChatAPIResponse
+
+from . import ops
+import asyncio
+import json
+from .executors import one_by_one
+
+ENV = 'LynxScribe'
+one_by_one.register(ENV)
+op = ops.op_registration(ENV)
+output_on_top = ops.output_position(output="top")
+
+@output_on_top
+@op("Vector store")
+def vector_store(*, name='chromadb', collection_name='lynx'):
+  vector_store = get_vector_store(name=name, collection_name=collection_name)
+  return {'vector_store': vector_store}
+
+@output_on_top
+@op("LLM")
+def llm(*, name='openai'):
+  llm = get_llm_engine(name=name)
+  return {'llm': llm}
+
+@output_on_top
+@ops.input_position(llm="bottom")
+@op("Text embedder")
+def text_embedder(llm, *, model='text-embedding-ada-002'):
+  llm = llm[0]['llm']
+  text_embedder = TextEmbedder(llm=llm, model=model)
+  return {'text_embedder': text_embedder}
+
+@output_on_top
+@ops.input_position(vector_store="bottom", text_embedder="bottom")
+@op("RAG graph")
+def rag_graph(vector_store, text_embedder):
+    vector_store = vector_store[0]['vector_store']
+    text_embedder = text_embedder[0]['text_embedder']
+    rag_graph = RAGGraph(
+        PandasKnowledgeBaseGraph(vector_store=vector_store, text_embedder=text_embedder)
+    )
+    return {'rag_graph': rag_graph}
+
+@output_on_top
+@op("Scenario selector")
+def scenario_selector(*, scenario_file: str, node_types='intent_cluster'):
+  scenarios = load_config(scenario_file)
+  node_types = [t.strip() for t in node_types.split(',')]
+  scenario_selector = ScenarioSelector(
+      scenarios=[Scenario(**scenario) for scenario in scenarios],
+      node_types=node_types,
+  )
+  return {'scenario_selector': scenario_selector}
+
+DEFAULT_NEGATIVE_ANSWER = "I'm sorry, but the data I've been trained on does not contain any information related to your question."
+
+@output_on_top
+@ops.input_position(rag_graph="bottom", scenario_selector="bottom", llm="bottom")
+@op("RAG chatbot")
+def rag_chatbot(
+    rag_graph, scenario_selector, llm, *,
+    negative_answer=DEFAULT_NEGATIVE_ANSWER,
+    limits_by_type='{}',
+    strict_limits=True, max_results=5):
+  rag_graph = rag_graph[0]['rag_graph']
+  scenario_selector = scenario_selector[0]['scenario_selector']
+  llm = llm[0]['llm']
+  limits_by_type = json.loads(limits_by_type)
+  rag_chatbot = RAGChatbot(
+      rag_graph=rag_graph,
+      scenario_selector=scenario_selector,
+      llm=llm,
+      negative_answer=negative_answer,
+      limits_by_type=limits_by_type,
+      strict_limits=strict_limits,
+      max_results=max_results,
+  )
+  return {'chatbot': rag_chatbot}
+
+@output_on_top
+@ops.input_position(processor="bottom")
+@op("Chat processor")
+def chat_processor(processor, *, _ctx: one_by_one.Context):
+  cfg = _ctx.last_result or {'question_processors': [], 'answer_processors': [], 'masks': []}
+  for f in ['question_processor', 'answer_processor', 'mask']:
+    if f in processor:
+      cfg[f + 's'].append(processor[f])
+  question_processors = cfg['question_processors'][:]
+  answer_processors = cfg['answer_processors'][:]
+  masking_templates = {}
+  for mask in cfg['masks']:
+    masking_templates[mask['name']] = mask
+  if masking_templates:
+    question_processors.append(MaskTemplate(masking_templates=masking_templates))
+    answer_processors.append(MaskTemplate(masking_templates=masking_templates))
+  chat_processor = ChatProcessor(question_processors=question_processors, answer_processors=answer_processors)
+  return {'chat_processor': chat_processor, **cfg}
+
+@output_on_top
+@op("Truncate history")
+def truncate_history(*, max_tokens=10000, language='English'):
+  return {'question_processor': TruncateHistory(max_tokens=max_tokens, language=language.lower())}
+
+@output_on_top
+@op("Mask")
+def mask(*, name='', regex='', exceptions='', mask_pattern=''):
+  exceptions = [e.strip() for e in exceptions.split(',') if e.strip()]
+  return {'mask': {'name': name, 'regex': regex, 'exceptions': exceptions, 'mask_pattern': mask_pattern}}
+
+@ops.input_position(chat_api="bottom")
+@op("Test Chat API")
+def test_chat_api(message, chat_api):
+  chat_api = chat_api[0]['chat_api']
+  request = ChatAPIRequest(session_id="b43215a0-428f-11ef-9454-0242ac120002", question=message['text'], history=[])
+  response = asyncio.run(chat_api.answer(request))
+  return {'response': response.answer}
+
+@op("Input chat")
+def input_chat(*, chat: str):
+  return {'text': chat}
+
+@output_on_top
+@ops.input_position(chatbot="bottom", chat_processor="bottom", knowledge_base="bottom")
+@op("Chat API")
+def chat_api(chatbot, chat_processor, knowledge_base, *, model='gpt-4o-mini'):
+  chatbot = chatbot[0]['chatbot']
+  chat_processor = chat_processor[0]['chat_processor']
+  knowledge_base = knowledge_base[0]
+  c = ChatAPI(
+      chatbot=chatbot,
+      chat_processor=chat_processor,
+      model=model,
+  )
+  if knowledge_base:
+    c.chatbot.rag_graph.kg_base.load_v1_knowledge_base(**knowledge_base)
+    c.chatbot.scenario_selector.check_compatibility(c.chatbot.rag_graph)
+  return {'chat_api': c}
+
+@output_on_top
+@op("Knowledge base")
+def knowledge_base(*, nodes_path='nodes.pickle', edges_path='edges.pickle', template_cluster_path='tempclusters.pickle'):
+    return {'nodes_path': nodes_path, 'edges_path': edges_path, 'template_cluster_path': template_cluster_path}
+
+@op("View", view="table_view")
+def view(input):
+  columns = [str(c) for c in input.keys() if not str(c).startswith('_')]
+  v = {
+    'dataframes': { 'df': {
+      'columns': columns,
+      'data': [[input[c] for c in columns]],
+    }}
+  }
+  return v
+
+async def api_service(request):
+  '''
+  Serves a chat endpoint that matches LynxScribe's interface.
+  To access it you need to add the "module" and "workspace"
+  parameters.
+  The workspace must contain exactly one "Chat API" node.
+
+    curl -X POST ${LYNXKITE_URL}/api/service \
+      -H "Content-Type: application/json" \
+      -d '{
+        "module": "server.lynxscribe_ops",
+        "workspace": "LynxScribe demo",
+        "session_id": "b43215a0-428f-11ef-9454-0242ac120002",
+        "question": "what does the fox say",
+        "history": [],
+        "user_id": "x",
+        "meta_inputs": {}
+      }'
+  '''
+  import pathlib
+  from . import workspace
+  DATA_PATH = pathlib.Path.cwd() / 'data'
+  path = DATA_PATH / request['workspace']
+  assert path.is_relative_to(DATA_PATH)
+  assert path.exists(), f'Workspace {path} does not exist'
+  ws = workspace.load(path)
+  contexts = ops.EXECUTORS[ENV](ws)
+  nodes = [op for op in ws.nodes if op.data.title == 'Chat API']
+  [node] = nodes
+  context = contexts[node.id]
+  chat_api = context.last_result['chat_api']
+  request = ChatAPIRequest(session_id=request['session_id'], question=request['question'], history=request['history'])
+  response = await chat_api.answer(request)
+  return response

server/main.py

@@ -1,15 +1,18 @@
 import dataclasses
 import fastapi
+import importlib
 import pathlib
 import pkgutil
 from . import ops
 from . import workspace
 
 here = pathlib.Path(__file__).parent
+lynxkite_modules = {}
 for _, name, _ in pkgutil.iter_modules([str(here)]):
     if name.endswith('_ops') and not name.startswith('test_'):
         print(f'Importing {name}')
-        __import__(f'server.{name}')
+        name = f'server.{name}'
+        lynxkite_modules[name] = importlib.import_module(name)
 
 app = fastapi.FastAPI()
 
@@ -67,3 +70,9 @@ def make_dir(req: dict):
     assert not path.exists()
     path.mkdir()
     return list_dir(path.parent)
+
+@app.post("/api/service")
+async def service(req: dict):
+    '''Executors can provide extra HTTP APIs through the /api/service endpoint.'''
+    module = lynxkite_modules[req['module']]
+    return await module.api_service(req)

web/package-lock.json

@@ -25,6 +25,7 @@
         "sass": "^1.77.2",
         "svelte": "^4.2.12",
         "svelte-check": "^3.6.9",
+        "svelte-markdown": "^0.4.1",
         "tslib": "^2.6.2",
         "typescript": "^5.4.4",
         "unplugin-icons": "^0.18.5",
@@ -1081,6 +1082,13 @@
       "resolved": "https://registry.npmjs.org/@types/geojson/-/geojson-7946.0.14.tgz",
       "integrity": "sha512-WCfD5Ht3ZesJUsONdhvm84dmzWOiOzOAqOncN0++w0lBw1o8OuDNJF2McvvCef/yBqb/HYRahp1BYtODFQ8bRg=="
     },
+    "node_modules/@types/marked": {
+      "version": "5.0.2",
+      "resolved": "https://registry.npmjs.org/@types/marked/-/marked-5.0.2.tgz",
+      "integrity": "sha512-OucS4KMHhFzhz27KxmWg7J+kIYqyqoW5kdIEI319hqARQQUTqhao3M/F+uFnDXD0Rg72iDDZxZNxq5gvctmLlg==",
+      "dev": true,
+      "license": "MIT"
+    },
     "node_modules/@types/pug": {
       "version": "2.0.10",
       "resolved": "https://registry.npmjs.org/@types/pug/-/pug-2.0.10.tgz",
@@ -1844,6 +1852,19 @@
         "@jridgewell/sourcemap-codec": "^1.4.15"
       }
     },
+    "node_modules/marked": {
+      "version": "5.1.2",
+      "resolved": "https://registry.npmjs.org/marked/-/marked-5.1.2.tgz",
+      "integrity": "sha512-ahRPGXJpjMjwSOlBoTMZAK7ATXkli5qCPxZ21TG44rx1KEo44bii4ekgTDQPNRQ4Kh7JMb9Ub1PVk1NxRSsorg==",
+      "dev": true,
+      "license": "MIT",
+      "bin": {
+        "marked": "bin/marked.js"
+      },
+      "engines": {
+        "node": ">= 16"
+      }
+    },
     "node_modules/mdn-data": {
       "version": "2.0.30",
       "resolved": "https://registry.npmjs.org/mdn-data/-/mdn-data-2.0.30.tgz",
@@ -2460,6 +2481,20 @@
         "svelte": "^3.19.0 || ^4.0.0"
       }
     },
+    "node_modules/svelte-markdown": {
+      "version": "0.4.1",
+      "resolved": "https://registry.npmjs.org/svelte-markdown/-/svelte-markdown-0.4.1.tgz",
+      "integrity": "sha512-pOlLY6EruKJaWI9my/2bKX8PdTeP5CM0s4VMmwmC2prlOkjAf+AOmTM4wW/l19Y6WZ87YmP8+ZCJCCwBChWjYw==",
+      "dev": true,
+      "license": "MIT",
+      "dependencies": {
+        "@types/marked": "^5.0.1",
+        "marked": "^5.1.2"
+      },
+      "peerDependencies": {
+        "svelte": "^4.0.0"
+      }
+    },
     "node_modules/svelte-preprocess": {
       "version": "5.1.4",
       "resolved": "https://registry.npmjs.org/svelte-preprocess/-/svelte-preprocess-5.1.4.tgz",

web/package.json

@@ -15,6 +15,7 @@
     "sass": "^1.77.2",
     "svelte": "^4.2.12",
     "svelte-check": "^3.6.9",
+    "svelte-markdown": "^0.4.1",
     "tslib": "^2.6.2",
     "typescript": "^5.4.4",
     "unplugin-icons": "^0.18.5",