Feature/seq agents (#2798)

* update build functions

* sequential agents

* update langchain to 0.2, added sequential agent nodes

* add marketplace templates

* update howto wordings

* Merge branch 'main' into feature/Seq-Agents

# Conflicts:
#	pnpm-lock.yaml

* update deprecated functions and add new sequential nodes

* add marketplace templates

* update marketplace templates, add structured output to llm node

* add multi agents template

* update llm node with bindmodels

* update cypress version

* update templates sticky note wordings

* update tool node to include human in loop action

* update structured outputs error from models

* update cohere package to resolve google genai pipeThrough bug

* update mistral package version, added message reconstruction before invoke seq agent

* add HITL to agent

* update state messages restructuring

* update load and split methods for s3 directory
This commit is contained in:
Henry Heng
2024-07-22 17:46:14 +01:00
committed by GitHub
parent 34d0e4302c
commit bca4de0c63
152 changed files with 55307 additions and 35236 deletions
@@ -1,7 +1,7 @@
import { omit } from 'lodash'
import { ICommonObject, INode, INodeData, INodeParams } from '../../../src/Interface'
import { ICommonObject, IDocument, INode, INodeData, INodeParams } from '../../../src/Interface'
import { TextSplitter } from 'langchain/text_splitter'
import { GithubRepoLoader, GithubRepoLoaderParams } from 'langchain/document_loaders/web/github'
import { GithubRepoLoader, GithubRepoLoaderParams } from '@langchain/community/document_loaders/web/github'
import { getCredentialData, getCredentialParam } from '../../../src'
class Github_DocumentLoaders implements INode {
@@ -139,7 +139,15 @@ class Github_DocumentLoaders implements INode {
if (ignorePath) githubOptions.ignorePaths = JSON.parse(ignorePath)
const loader = new GithubRepoLoader(repoLink, githubOptions)
let docs = textSplitter ? await loader.loadAndSplit(textSplitter) : await loader.load()
let docs: IDocument[] = []
if (textSplitter) {
docs = await loader.load()
docs = await textSplitter.splitDocuments(docs)
} else {
docs = await loader.load()
}
if (metadata) {
const parsedMetadata = typeof metadata === 'object' ? metadata : JSON.parse(metadata)