Run Open WebUI against LynxScribe in LynxKite.

README.md

@@ -7,16 +7,13 @@ original LynxKite. The primary goals of this rewrite are:
 - More extensible backend. Make it easy to add new LynxKite boxes. Make it easy to use our frontend for other purposes,
   configuring and executing other pipelines.
 
-Current status: **PROTOTYPE**
-
 ## Installation
 
 To run the backend:
 
 ```bash
-pip install -r requirements.txt
-PYTHONPATH=. pydantic2ts --module server.workspace --output ./web/src/apiTypes.ts
-uvicorn server.main:app --reload
+PYTHONPATH=. uv run pydantic2ts --module server.workspace --output ./web/src/apiTypes.ts --json2ts-cmd "npm exec --prefix web json2ts"
+uv run fastapi run server/main.py --reload
 ```
 
 To run the frontend:
@@ -26,3 +23,9 @@ cd web
 npm i
 npm run dev
 ```
+
+To run a chat UI for LynxScribe workspaces:
+
+```bash
+WEBUI_AUTH=false OPENAI_API_BASE_URL=http://localhost:8000/api/service/server.lynxscribe_ops uvx open-webui serve
+```

data/LynxScribe demo

@@ -580,8 +580,7 @@
       "data": {
         "title": "Truncate history",
         "params": {
-          "max_tokens": 10000,
-          "language": "English"
+          "max_tokens": 10000
         },
         "display": null,
         "error": null,
@@ -594,13 +593,6 @@
               "type": {
                 "type": "<class 'int'>"
               }
-            },
-            "language": {
-              "name": "language",
-              "default": "English",
-              "type": {
-                "type": "<class 'str'>"
-              }
             }
           },
           "inputs": {},
@@ -957,4 +949,4 @@
       "targetHandle": "chat_api"
     }
   ]
-}
+}

server/executors/one_by_one.py

@@ -127,12 +127,13 @@ async def execute(ws, catalog, cache=None):
             results = []
             for task in ts:
                 try:
-                    inputs = [
-                        batch_inputs[(n, i.name)]
-                        if i.position in "top or bottom"
-                        else task
-                        for i in op.inputs.values()
-                    ]
+                    inputs = []
+                    for i in op.inputs.values():
+                        if i.position in "top or bottom":
+                            assert (n, i.name) in batch_inputs, f"{i.name} is missing"
+                            inputs.append(batch_inputs[(n, i.name)])
+                        else:
+                            inputs.append(task)
                     if cache is not None:
                         key = make_cache_key((inputs, params))
                         if key not in cache:

server/llm_ops.py

@@ -13,7 +13,7 @@ from .executors import one_by_one
 chat_client = openai.OpenAI(base_url="http://localhost:8080/v1")
 embedding_client = openai.OpenAI(base_url="http://localhost:7997/")
 jinja = jinja2.Environment()
-chroma_client = chromadb.Client()
+chroma_client = None
 LLM_CACHE = {}
 ENV = "LLM logic"
 one_by_one.register(ENV)
@@ -178,12 +178,15 @@ def rag(
     num_matches: int = 10,
     _ctx: one_by_one.Context,
 ):
+    global chroma_client
     if engine == RagEngine.Chroma:
         last = _ctx.last_result
         if last:
             collection = last["_collection"]
         else:
             collection_name = _ctx.node.id.replace(" ", "_")
+            if chroma_client is None:
+                chroma_client = chromadb.Client()
             for c in chroma_client.list_collections():
                 if c.name == collection_name:
                     chroma_client.delete_collection(name=collection_name)

server/lynxkite_ops.py

@@ -218,7 +218,7 @@ def sql(bundle: Bundle, *, query: ops.LongStr, save_as: str = "result"):
             # TODO: Currently `collect()` moves the data from cuDF to Polars. Then we convert it to Pandas,
             # which (hopefully) puts it back into cuDF. Hopefully we will be able to keep it in cuDF.
     else:
-        res = pl.SQLContext(bundle.dfs).execute(query)
+        res = pl.SQLContext(bundle.dfs).execute(query).collect().to_pandas()
     bundle.dfs[save_as] = res
     return bundle
 

server/lynxscribe_ops.py

@@ -5,19 +5,18 @@ 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.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 (
+from lynxscribe.components.chat.processors import (
+    ChatProcessor,
     MaskTemplate,
     TruncateHistory,
 )
-from lynxscribe.components.chat_api import ChatAPI, ChatAPIRequest, ChatAPIResponse
+from lynxscribe.components.chat.api import ChatAPI, ChatAPIRequest, ChatAPIResponse
 
 from . import ops
-import asyncio
 import json
 from .executors import one_by_one
 
@@ -134,12 +133,8 @@ def chat_processor(processor, *, _ctx: one_by_one.Context):
 
 @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()
-        )
-    }
+def truncate_history(*, max_tokens=10000):
+    return {"question_processor": TruncateHistory(max_tokens=max_tokens)}
 
 
 @output_on_top
@@ -224,42 +219,102 @@ def view(input):
     return v
 
 
-async def api_service(request):
+async def get_chat_api(ws):
+    import pathlib
+    from . import workspace
+
+    DATA_PATH = pathlib.Path.cwd() / "data"
+    path = DATA_PATH / ws
+    assert path.is_relative_to(DATA_PATH)
+    assert path.exists(), f"Workspace {path} does not exist"
+    ws = workspace.load(path)
+    contexts = await ops.EXECUTORS[ENV](ws)
+    nodes = [op for op in ws.nodes if op.data.title == "Chat API"]
+    [node] = nodes
+    context = contexts[node.id]
+    return context.last_result["chat_api"]
+
+
+async def stream_chat_api_response(request):
+    chat_api = await get_chat_api(request["model"])
+    chat_api_request = ChatAPIRequest(
+        session_id=request.get("session_id", "00000000-0000-0000-0000-000000000000"),
+        history=request["messages"][:-1],
+        question=request["messages"][-1]["content"],
+    )
+    response = await chat_api.answer(chat_api_request)
+    response = response.model_dump()
+    yield json.dumps(
+        {
+            **response,
+            "id": "asd",
+            "object": "chat.completion.chunk",
+            "model": request["model"],
+            "choices": [
+                {
+                    "index": 0,
+                    "delta": {"role": "assistant", "content": response["answer"]},
+                }
+            ],
+        }
+    )
+
+
+async def api_service_post(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 \
+      curl -X POST ${LYNXKITE_URL}/api/service/server.lynxkite_ops \
         -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": {}
+          "model": "LynxScribe demo",
+          "messages": [{"role": "user", "content": "what does the fox say"}]
         }'
-  """
+    """
+    path = "/".join(request.url.path.split("/")[4:])
+    request = await request.json()
+    if path == "chat/completions":
+        from sse_starlette.sse import EventSourceResponse
+
+        return EventSourceResponse(stream_chat_api_response(request))
+    return {"error": "Not found"}
+
+
+async def api_service_get(request):
+    path = "/".join(request.url.path.split("/")[4:])
+    if path == "models":
+        return {
+            "object": "list",
+            "data": [
+                {
+                    "id": ws,
+                    "object": "model",
+                    "created": 0,
+                    "owned_by": "lynxkite",
+                    "meta": {"profile_image_url": "https://lynxkite.com/favicon.png"},
+                }
+                for ws in get_lynxscribe_workspaces()
+            ],
+        }
+    return {"error": "Not found"}
+
+
+def get_lynxscribe_workspaces():
     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
+    DATA_DIR = pathlib.Path.cwd() / "data"
+    workspaces = []
+    for p in DATA_DIR.glob("**/*"):
+        if p.is_file():
+            try:
+                ws = workspace.load(p)
+                if ws.env == ENV:
+                    workspaces.append(p.relative_to(DATA_DIR))
+            except Exception:
+                pass  # Ignore files that are not valid workspaces.
+    workspaces.sort()
+    return workspaces

server/main.py

@@ -1,3 +1,6 @@
+# TODO: Make this conditional. Until then just comment/uncomment it to use cuDF Pandas.
+# import cudf.pandas
+# cudf.pandas.install()
 import dataclasses
 import fastapi
 import importlib
@@ -87,8 +90,15 @@ def make_dir(req: dict):
     return list_dir(path.parent)
 
 
-@app.post("/api/service")
-async def service(req: dict):
+@app.get("/api/service/{module_path:path}")
+async def service_get(req: fastapi.Request, module_path: str):
     """Executors can provide extra HTTP APIs through the /api/service endpoint."""
-    module = lynxkite_modules[req["module"]]
-    return await module.api_service(req)
+    module = lynxkite_modules[module_path.split("/")[0]]
+    return await module.api_service_get(req)
+
+
+@app.post("/api/service/{module_path:path}")
+async def service_post(req: fastapi.Request, module_path: str):
+    """Executors can provide extra HTTP APIs through the /api/service endpoint."""
+    module = lynxkite_modules[module_path.split("/")[0]]
+    return await module.api_service_post(req)

web/package-lock.json

@@ -20,6 +20,7 @@
         "daisyui": "^4.12.20",
         "echarts": "^5.5.1",
         "fuse.js": "^7.0.0",
+        "json-schema-to-typescript": "^15.0.3",
         "react": "^18.3.1",
         "react-dom": "^18.3.1",
         "react-markdown": "^9.0.1",
@@ -97,6 +98,23 @@
         "url": "https://github.com/sponsors/antfu"
       }
     },
+    "node_modules/@apidevtools/json-schema-ref-parser": {
+      "version": "11.7.3",
+      "resolved": "https://registry.npmjs.org/@apidevtools/json-schema-ref-parser/-/json-schema-ref-parser-11.7.3.tgz",
+      "integrity": "sha512-WApSdLdXEBb/1FUPca2lteASewEfpjEYJ8oXZP+0gExK5qSfsEKBKcA+WjY6Q4wvXwyv0+W6Kvc372pSceib9w==",
+      "license": "MIT",
+      "dependencies": {
+        "@jsdevtools/ono": "^7.1.3",
+        "@types/json-schema": "^7.0.15",
+        "js-yaml": "^4.1.0"
+      },
+      "engines": {
+        "node": ">= 16"
+      },
+      "funding": {
+        "url": "https://github.com/sponsors/philsturgeon"
+      }
+    },
     "node_modules/@babel/code-frame": {
       "version": "7.26.2",
       "resolved": "https://registry.npmjs.org/@babel/code-frame/-/code-frame-7.26.2.tgz",
@@ -687,6 +705,12 @@
         "@jridgewell/sourcemap-codec": "^1.4.14"
       }
     },
+    "node_modules/@jsdevtools/ono": {
+      "version": "7.1.3",
+      "resolved": "https://registry.npmjs.org/@jsdevtools/ono/-/ono-7.1.3.tgz",
+      "integrity": "sha512-4JQNk+3mVzK3xh2rqd6RB4J46qUR19azEHBneZyTZM+c456qOrbbM/5xcR8huNCCcbVt7+UmizG6GuUvPvKUYg==",
+      "license": "MIT"
+    },
     "node_modules/@nodelib/fs.scandir": {
       "version": "2.1.5",
       "resolved": "https://registry.npmjs.org/@nodelib/fs.scandir/-/fs.scandir-2.1.5.tgz",
@@ -1333,7 +1357,12 @@
       "version": "7.0.15",
       "resolved": "https://registry.npmjs.org/@types/json-schema/-/json-schema-7.0.15.tgz",
       "integrity": "sha512-5+fP8P8MFNC+AyZCDxrB2pkZFPGzqQWUzpSeuuVLvm8VMcorNYavBqoFcxK8bQz4Qsbn4oUEEem4wDLfcysGHA==",
-      "dev": true,
+      "license": "MIT"
+    },
+    "node_modules/@types/lodash": {
+      "version": "4.17.14",
+      "resolved": "https://registry.npmjs.org/@types/lodash/-/lodash-4.17.14.tgz",
+      "integrity": "sha512-jsxagdikDiDBeIRaPYtArcT8my4tN1og7MtMRquFT3XNA6axxyHDRUemqDz/taRDdOUn0GnGHRCuff4q48sW9A==",
       "license": "MIT"
     },
     "node_modules/@types/mdast": {
@@ -3339,7 +3368,6 @@
       "version": "2.1.1",
       "resolved": "https://registry.npmjs.org/is-extglob/-/is-extglob-2.1.1.tgz",
       "integrity": "sha512-SbKbANkN603Vi4jEZv49LeVJMn4yGwsbzZworEoyEiutsN3nJYdbO36zfhGJ6QEDpOZIFkDtnq5JRxmvl3jsoQ==",
-      "dev": true,
       "license": "MIT",
       "engines": {
         "node": ">=0.10.0"
@@ -3359,7 +3387,6 @@
       "version": "4.0.3",
       "resolved": "https://registry.npmjs.org/is-glob/-/is-glob-4.0.3.tgz",
       "integrity": "sha512-xelSayHH36ZgE7ZWhli7pW34hNbNl8Ojv5KVmkJD4hBdD3th8Tfk9vYasLM+mXWOZhFkgZfxhLSnrwRr4elSSg==",
-      "dev": true,
       "license": "MIT",
       "dependencies": {
         "is-extglob": "^2.1.1"
@@ -3486,6 +3513,29 @@
       "integrity": "sha512-xyFwyhro/JEof6Ghe2iz2NcXoj2sloNsWr/XsERDK/oiPCfaNhl5ONfp+jQdAZRQQ0IJWNzH9zIZF7li91kh2w==",
       "license": "MIT"
     },
+    "node_modules/json-schema-to-typescript": {
+      "version": "15.0.3",
+      "resolved": "https://registry.npmjs.org/json-schema-to-typescript/-/json-schema-to-typescript-15.0.3.tgz",
+      "integrity": "sha512-iOKdzTUWEVM4nlxpFudFsWyUiu/Jakkga4OZPEt7CGoSEsAsUgdOZqR6pcgx2STBek9Gm4hcarJpXSzIvZ/hKA==",
+      "license": "MIT",
+      "dependencies": {
+        "@apidevtools/json-schema-ref-parser": "^11.5.5",
+        "@types/json-schema": "^7.0.15",
+        "@types/lodash": "^4.17.7",
+        "is-glob": "^4.0.3",
+        "js-yaml": "^4.1.0",
+        "lodash": "^4.17.21",
+        "minimist": "^1.2.8",
+        "prettier": "^3.2.5",
+        "tinyglobby": "^0.2.9"
+      },
+      "bin": {
+        "json2ts": "dist/src/cli.js"
+      },
+      "engines": {
+        "node": ">=16.0.0"
+      }
+    },
     "node_modules/json-schema-traverse": {
       "version": "0.4.1",
       "resolved": "https://registry.npmjs.org/json-schema-traverse/-/json-schema-traverse-0.4.1.tgz",
@@ -3760,6 +3810,12 @@
         "url": "https://github.com/sponsors/sindresorhus"
       }
     },
+    "node_modules/lodash": {
+      "version": "4.17.21",
+      "resolved": "https://registry.npmjs.org/lodash/-/lodash-4.17.21.tgz",
+      "integrity": "sha512-v2kDEe57lecTulaDIuNTPy3Ry4gLGJ6Z1O3vE1krgXZNrsQ+LFTGHVxVjcXPs17LhbZVGedAJv8XZ1tvj5FvSg==",
+      "license": "MIT"
+    },
     "node_modules/lodash.debounce": {
       "version": "4.0.8",
       "resolved": "https://registry.npmjs.org/lodash.debounce/-/lodash.debounce-4.0.8.tgz",
@@ -4452,6 +4508,15 @@
         "node": "*"
       }
     },
+    "node_modules/minimist": {
+      "version": "1.2.8",
+      "resolved": "https://registry.npmjs.org/minimist/-/minimist-1.2.8.tgz",
+      "integrity": "sha512-2yyAR8qBkN3YuheJanUpWC5U3bb5osDywNB8RzDVlDwDHbocAJveqqj1u8+SVD7jkWT4yvsHCpWqqWqAxb0zCA==",
+      "license": "MIT",
+      "funding": {
+        "url": "https://github.com/sponsors/ljharb"
+      }
+    },
     "node_modules/minipass": {
       "version": "7.1.2",
       "resolved": "https://registry.npmjs.org/minipass/-/minipass-7.1.2.tgz",
@@ -4985,6 +5050,21 @@
         "node": ">= 0.8.0"
       }
     },
+    "node_modules/prettier": {
+      "version": "3.4.2",
+      "resolved": "https://registry.npmjs.org/prettier/-/prettier-3.4.2.tgz",
+      "integrity": "sha512-e9MewbtFo+Fevyuxn/4rrcDAaq0IYxPGLvObpQjiZBMAzB9IGmzlnG9RZy3FFas+eBMu2vA0CszMeduow5dIuQ==",
+      "license": "MIT",
+      "bin": {
+        "prettier": "bin/prettier.cjs"
+      },
+      "engines": {
+        "node": ">=14"
+      },
+      "funding": {
+        "url": "https://github.com/prettier/prettier?sponsor=1"
+      }
+    },
     "node_modules/property-information": {
       "version": "6.5.0",
       "resolved": "https://registry.npmjs.org/property-information/-/property-information-6.5.0.tgz",
@@ -5695,6 +5775,45 @@
       "integrity": "sha512-WiCJLEECkO18gwqIp6+hJg0//p23HXp4S+gGtAKu3mI2F2/sXC4FvHvXvB0zJVVaTPhx1/tOwdbRsa1sOBIKqQ==",
       "license": "MIT"
     },
+    "node_modules/tinyglobby": {
+      "version": "0.2.10",
+      "resolved": "https://registry.npmjs.org/tinyglobby/-/tinyglobby-0.2.10.tgz",
+      "integrity": "sha512-Zc+8eJlFMvgatPZTl6A9L/yht8QqdmUNtURHaKZLmKBE12hNPSrqNkUp2cs3M/UKmNVVAMFQYSjYIVHDjW5zew==",
+      "license": "MIT",
+      "dependencies": {
+        "fdir": "^6.4.2",
+        "picomatch": "^4.0.2"
+      },
+      "engines": {
+        "node": ">=12.0.0"
+      }
+    },
+    "node_modules/tinyglobby/node_modules/fdir": {
+      "version": "6.4.2",
+      "resolved": "https://registry.npmjs.org/fdir/-/fdir-6.4.2.tgz",
+      "integrity": "sha512-KnhMXsKSPZlAhp7+IjUkRZKPb4fUyccpDrdFXbi4QL1qkmFh9kVY09Yox+n4MaOb3lHZ1Tv829C3oaaXoMYPDQ==",
+      "license": "MIT",
+      "peerDependencies": {
+        "picomatch": "^3 || ^4"
+      },
+      "peerDependenciesMeta": {
+        "picomatch": {
+          "optional": true
+        }
+      }
+    },
+    "node_modules/tinyglobby/node_modules/picomatch": {
+      "version": "4.0.2",
+      "resolved": "https://registry.npmjs.org/picomatch/-/picomatch-4.0.2.tgz",
+      "integrity": "sha512-M7BAV6Rlcy5u+m6oPhAPFgJTzAioX/6B0DxyvDlo9l8+T3nLKbrczg2WLUyzd45L8RqfUMyGPzekbMvX2Ldkwg==",
+      "license": "MIT",
+      "engines": {
+        "node": ">=12"
+      },
+      "funding": {
+        "url": "https://github.com/sponsors/jonschlinkert"
+      }
+    },
     "node_modules/to-regex-range": {
       "version": "5.0.1",
       "resolved": "https://registry.npmjs.org/to-regex-range/-/to-regex-range-5.0.1.tgz",

web/package.json

@@ -22,6 +22,7 @@
     "daisyui": "^4.12.20",
     "echarts": "^5.5.1",
     "fuse.js": "^7.0.0",
+    "json-schema-to-typescript": "^15.0.3",
     "react": "^18.3.1",
     "react-dom": "^18.3.1",
     "react-markdown": "^9.0.1",

web/src/apiTypes.ts

@@ -5,6 +5,13 @@
 /* Do not modify it by hand - just update the pydantic models and then re-run the script
 */
 
+/* eslint-disable */
+/**
+ * This file was automatically generated by json-schema-to-typescript.
+ * DO NOT MODIFY IT BY HAND. Instead, modify the source JSONSchema file,
+ * and run json-schema-to-typescript to regenerate this file.
+ */
+
 export interface BaseConfig {
   [k: string]: unknown;
 }