2.0.0 • Published 5 months ago

@graphai/agent_filters v2.0.0

Weekly downloads
-
License
MIT
Repository
github
Last release
5 months ago

@graphai/agent_filters for GraphAI

Agent filter collections for GraphAI.

Install

yarn add @graphai/agent_filters

Usage

namedInput Validator

Validate values of namedInputs based on the input schema of agentFunctionInfo

https://github.com/isamu/graphai/blob/agentFilter/packages/agent_filters/tests/validation/test_agent_namedinput_validator.ts

import { GraphAI } from "graphai";
import * as agents from "@graphai/agents";
import { namedInputValidatorFilter } from "@graphai/agent_filters";

const agentFilters = [
  {
    name: "namedInputValidatorFilter",
    agent: namedInputValidatorFilter,
  },
];

const graph = new GraphAI(graph_data, agents, { agentFilters });
const results = await graph.run();

streamAgentFilterGenerator

Receive stream data externally through streamTokenCallback of filterParams. Available for both server and client. See express code.

server

https://github.com/receptron/graphai_utils/blob/main/packages/express/src/express.ts

express server

    return async (req: express.Request, res: express.Response) => {
      res.setHeader("Content-Type", "text/event-stream;charset=utf-8");
      res.setHeader("Cache-Control", "no-cache, no-transform");
      res.setHeader("X-Accel-Buffering", "no");

      const callback = (context: AgentFunctionContext, token: string) => {
        if (token) {
          res.write(token);
        }
      };
      const streamAgentFilter = {
        name: "streamAgentFilter",
        agent: streamAgentFilterGenerator<string>(callback),
      };
      const agentFilters = [streamAgentFilter]

      const agentFilterRunner = agentFilterRunnerBuilder(agentFilters);
      const result = await agentFilterRunner(context, agent.agent);

      const json_data = JSON.stringify(result);
      res.write("___END___");
      res.write(json_data);
      return res.end();
   }

web client

https://github.com/isamu/graphai-stream-web/blob/main/src/views/Home.vue

const useAgentFilter = (callback: (context: AgentFunctionContext, data: T) => void) => {
  const streamAgentFilter = streamAgentFilterGenerator(callback);

  const agentFilters = [
    {
      name: "streamAgentFilter",
      agent: streamAgentFilter,
      agentIds: streamAgents,
    },
  ];
  return agentFilters;
};   

export default defineComponent({
  setup() {
    const streamingData = ref<Record<string, unknown>>({});

    const callback = (context: AgentFunctionContext, data: string) => {
      const { nodeId } = context.debugInfo;
      streamingData.value[nodeId] = (streamingData.value[nodeId] ?? "") + data;
    };
    const agentFilters = useAgentFilter(callback);
    
    const graphai = new GraphAI(graphData, agents, { agentFilters });
  }
})

httpAgentFilter

In graph flow, bypass the agent execution and run it on the server via http.

Refer to the web sample.

https://github.com/isamu/graphai-stream-web/blob/main/src/views/Home.vue

flowchart TD
    subgraph "Http Agent Filter"
        Node1((Node 1)) --> Node2((Node 2))
        Node1 --> Node3((Node 3))
        Node2 --> Node4((Node 4))
        Node3 --> Node4
        
        %% Node execution flow
        Node1 -..-> |execution| Filter1[Agent Filter]
        Node2 -..-> |execution| Filter2[Agent Filter]
        Node3 -..-> |execution| Filter3[Agent Filter]
        Node4 -..-> |execution| Filter4[Agent Filter]
        
        Filter1 -..-> |browser execution| BrowserExec1[Browser Execution]
        Filter2 -..-> |browser execution| BrowserExec2[Browser Execution]
        Filter3 -..-> |server execution| ServerExec1[Server HTTP Request]
        Filter4 -..-> |server execution| ServerExec2[Server HTTP Request]
        
        ServerExec1 -..-> Server1[Server Process]
        ServerExec2 -..-> Server2[Server Process]
        
        %% Execution results
        BrowserExec1 -..-> |result| Node1
        BrowserExec2 -..-> |result| Node2
        Server1 -..-> |result| Node3
        Server2 -..-> |result| Node4
    end
    
    classDef graphNode fill:#98fb98,stroke:#333,stroke-width:2px
    classDef filterNode fill:#a7c7e7,stroke:#333,stroke-width:1px
    classDef browserNode fill:#ffb6c1,stroke:#333,stroke-width:1px
    classDef serverNode fill:#ffcc99,stroke:#333,stroke-width:1px
    
    class Node1,Node2,Node3,Node4 graphNode
    class Filter1,Filter2,Filter3,Filter4 filterNode
    class BrowserExec1,BrowserExec2 browserNode
    class ServerExec1,ServerExec2,Server1,Server2 serverNode

agentFilterRunnerBuilder

Execute agent filter and agent in test for agent filter usage It also run agent filter and agent outside of GraphAI. This is useful in Express web server.

see express sample and test code.

express https://github.com/receptron/graphai_utils/blob/main/packages/express/src/express.ts

test https://github.com/isamu/graphai/blob/agentFilter/packages/agent_filters/tests/filters/test_filter_runner.ts

ConsoleStepRunner

To debug graph data in the console, you can execute each agent step by step.

You need to install the @inquirer/input npm package.

It can be used via:

import { consoleStepRunner } from "@graphai/agent_filters";

const agentFilters = [
  {
    name: "consoleStepRunner",
    agent: consoleStepRunner,
  },
];

const graph = new GraphAI(
  graph_data,
  {
    ...agents,
  },
  { agentFilters },
);
await graph.run();
1.0.2

5 months ago

1.0.1

7 months ago

1.0.0

8 months ago

1.0.7

5 months ago

1.0.6

5 months ago

1.0.5

5 months ago

1.0.4

5 months ago

1.0.3

5 months ago

0.2.1

9 months ago

0.2.0

9 months ago

0.0.9

10 months ago

0.0.8

11 months ago

0.2.2

9 months ago

0.0.7

11 months ago

2.0.0

5 months ago

0.0.6

1 year ago

0.0.5

1 year ago

0.0.4

1 year ago

0.0.3

1 year ago

0.0.2

1 year ago

0.0.1

1 year ago