feat: initial implementation

Signed-off-by: Matous Havlena <[email protected]>

.dockerignore

@@ -0,0 +1 @@
+node_modules

.env.example

@@ -0,0 +1,119 @@
+# Port used by the server
+PORT=4000 # optional
+
+# Log level, set to "trace" for detailed logs
+LOG_LEVEL=info # optional
+
+# Secret used by the server to encrypt/decrypt values
+# You can use `$ openssl rand -base64 32` to generate new one
+CRYPTO_CIPHER_KEY=
+
+# MongoDB database, must be a replica set, see https://www.mongodb.com/docs/manual/tutorial/deploy-replica-set/
+MONGODB_URL=mongodb://localhost:27017?directConnection=true
+MONGODB_DATABASE_NAME=bee-api
+MONGODB_CA_CERT= # optional
+
+# Redis database, used by distributed queue and as pub/sub broker
+REDIS_URL=redis://127.0.0.1:6379/0
+REDIS_CA_CERT= # optional
+
+# Redis database for data caching
+# In production, make sure you set up eviction policy to `volatile-lru`
+# https://redis.io/docs/latest/develop/reference/eviction/#eviction-policies
+REDIS_CACHE_URL=redis://127.0.0.1:6379/1
+REDIS_CACHE_CA_CERT= # optional
+
+# Object Storage used for files and their text extractions 
+S3_ENDPOINT=
+S3_ACCESS_KEY_ID=
+S3_SECRET_ACCESS_KEY=
+S3_BUCKET_FILE_STORAGE=
+
+# OAuth2 client credentials for access token validation, acting as the resource server
+AUTH_WELL_KNOWN=
+AUTH_CLIENT_ID=
+AUTH_CLIENT_SECRET=
+AUTH_AUDIENCE=
+
+# Queue workers
+RUN_BULLMQ_WORKERS=runs,runs:cleanup,vectorStores:cleanup,vectorStores:fileProcessor,files:extraction,threads:cleanup,files:cleanup
+
+
+# --- BACKEND SECTION ---
+
+# LLM backend, possible values are: ollama, openai, watsonx, bam, ibm-vllm
+LLM_BACKEND=watsonx
+
+# Embedding backend, possible values are: ollama, openai, watsonx, bam, caikit
+EMBEDDING_BACKEND=watsonx
+
+# --- Backend details (only applicable to backend(s) selected above) ---
+
+# https://ollama.com/
+OLLAMA_URL=
+
+# https://openai.com/
+OPENAI_API_KEY=
+
+# https://www.ibm.com/products/watsonx-ai
+WATSONX_API_KEY=
+WATSONX_PROJECT_ID=
+
+BAM_API_KEY=
+
+# Must contain port, can contain {model_id} placeholder, e.g. "{model_id}.inference.example.com:443"
+IBM_VLLM_URL=
+IBM_VLLM_ROOT_CERT=
+IBM_VLLM_CERT_CHAIN=
+IBM_VLLM_PRIVATE_KEY=
+
+CAIKIT_URL=
+CAIKIT_CA_CERT=
+CAIKIT_CERT=
+CAIKIT_KEY=
+
+# --- BACKEND SECTION ---
+
+# --- TOOLS SECTION ---
+
+# Code Interpreter, see https://github.com/i-am-bee/bee-code-interpreter
+BEE_CODE_INTERPRETER_URL=http://localhost:50051
+BEE_CODE_INTERPRETER_CA_CERT= # optional
+BEE_CODE_INTERPRETER_CERT= # optional
+BEE_CODE_INTERPRETER_KEY= # optional
+# Storage for code interpreter, possible values are: s3, filesystem
+BEE_CODE_INTERPRETER_STORAGE_BACKEND=s3
+# applicable for s3 storage, uses S3 credential above
+BEE_CODE_INTERPRETER_BUCKET_FILE_STORAGE=
+# applicable for filesystem storage
+BEE_CODE_INTERPRETER_FILE_STORAGE_PATH=
+
+# Milvus is used as a vector store for file search tool
+MILVUS_HOST=127.0.0.1
+MILVUS_PORT=19530
+MILVUS_USE_TLS=false
+MILVUS_USERNAME=
+MILVUS_PASSWORD=
+MILVUS_DATABASE_NAME=
+
+# Search tool backend, possible values are: google, duck-duck-go
+SEARCH_TOOL_BACKEND=duck-duck-go
+# applicable for google search tool
+BEE_GOOGLE_SEARCH_API_KEY=
+BEE_GOOGLE_SEARCH_CSE_ID=
+
+# Forward proxy for api calling tool
+HTTP_PROXY_URL= # optional
+
+# --- TOOLS SECTION ---
+
+# Observer server, see https://github.com/i-am-bee/bee-observe
+BEE_OBSERVE_API_URL= # optional
+BEE_OBSERVE_API_AUTH_KEY= # optional
+
+# Text extraction service
+WDU_URL= # optional
+
+# Identifiers to be used for seeder and to allocate new users into default org/project
+ORGANIZATION_OWNER_ID_DEFAULT=org_user_670cc04869ddffe24f4fd70e
+PROJECT_ADMIN_ID_DEFAULT=proj_principal_670cc04869ddffe24f4fd710

.env.test.example

@@ -0,0 +1,2 @@
+BEE_URL=http://localhost:4000
+BEE_TOKEN=

.eslintignore

@@ -0,0 +1 @@
+src/**/grpc/types/**

.eslintrc.cjs

@@ -0,0 +1,44 @@
+module.exports = {
+  root: true,
+  parser: '@typescript-eslint/parser',
+  extends: [
+    'eslint:recommended',
+    'plugin:@typescript-eslint/recommended',
+    'plugin:import/recommended',
+    'plugin:import/typescript',
+    'plugin:prettier/recommended'
+  ],
+  parserOptions: {
+    project: ['tsconfig.json'],
+    tsconfigRootDir: __dirname
+  },
+  plugins: ['@typescript-eslint', 'import'],
+  settings: {
+    'import/resolver': {
+      typescript: true,
+      node: true
+    }
+  },
+  rules: {
+    'import/order': [
+      'error',
+      {
+        'newlines-between': 'always'
+      }
+    ],
+    'import/default': 'off',
+    '@typescript-eslint/no-explicit-any': 'off',
+    'import/no-named-as-default-member': 'off',
+    'import/no-named-as-default': 'off',
+    '@typescript-eslint/no-unused-vars': [
+      'error',
+      {
+        argsIgnorePattern: '^_',
+        varsIgnorePattern: '^_',
+        caughtErrorsIgnorePattern: '^_',
+      },
+    ],
+    '@typescript-eslint/no-empty-function': 'off',
+    'no-console': 'error'
+  }
+};

.gitignore

@@ -0,0 +1,48 @@
+# See https://help.github.com/articles/ignoring-files/ for more about ignoring files.
+
+# dependencies
+/node_modules
+/.pnp
+.pnp.js
+.yarn/install-state.gz
+
+# testing
+/coverage
+
+# production
+/dist
+
+# playground
+/playground
+
+# misc
+.DS_Store
+*.pem
+
+# debug
+npm-debug.log*
+yarn-debug.log*
+yarn-error.log*
+
+# env files
+.env
+.env.test
+
+# typescript
+*.tsbuildinfo
+
+# vscode
+.vscode
+
+# mikro-orm ts metadata cache
+/temp
+
+# npm
+.npmrc
+
+# other
+/tmp
+
+
+# WebStorm
+.idea

.husky/pre-commit

@@ -0,0 +1,2 @@
+pnpm exec lint-staged
+CI=true pnpm copyright

.nvmrc

@@ -0,0 +1 @@
+v22.2.0

.prettierrc

@@ -0,0 +1,5 @@
+{
+  "printWidth": 100,
+  "singleQuote": true,
+  "trailingComma": "none"
+}

Dockerfile

@@ -0,0 +1,54 @@
+FROM node:22.2-alpine AS base
+
+# Check https://github.com/nodejs/docker-node/tree/b4117f9333da4138b03a546ec926ef50a31506c3#nodealpine to understand why libc6-compat might be needed.
+RUN apk add --no-cache libc6-compat
+
+ENV APP_DIR=/app
+
+RUN mkdir -p ${APP_DIR}
+WORKDIR ${APP_DIR}
+
+# Install dependencies only when needed
+FROM base AS deps
+
+ENV CI=1
+ENV HUSKY=0
+
+# Install dependencies based on the preferred package manager
+COPY package.json pnpm-lock.yaml* ./
+RUN corepack enable pnpm && pnpm i --frozen-lockfile
+
+# Rebuild the source code only when needed
+FROM base AS builder
+
+ENV CI=1
+
+COPY --from=deps ${APP_DIR}/node_modules ./node_modules
+COPY . .
+
+RUN corepack enable pnpm && pnpm run build;
+
+# Production image, copy all the files and run next
+FROM base AS runner
+
+ENV NODE_ENV production
+
+# https://github.com/nodejs/docker-node/blob/main/docs/BestPractices.md
+RUN deluser --remove-home node \
+  && addgroup -S node -g 1001 \
+  && adduser -S -G node -u 1001 node
+
+COPY --chown=node:node --from=deps ${APP_DIR}/package.json ./
+COPY --chown=node:node --from=deps ${APP_DIR}/node_modules ./node_modules
+COPY --chown=node:node --from=builder ${APP_DIR}/dist ./dist
+
+RUN chown -R 1001:0 ${APP_DIR} &&\
+  chmod -R g+w ${APP_DIR}
+
+USER node
+
+EXPOSE 4000
+
+ENV PORT 4000
+
+CMD [ "node", "--enable-source-maps", "--experimental-loader=@opentelemetry/instrumentation/hook.mjs", "--import", "./dist/opentelemetry.js", "./dist/server.js" ]

README.md

@@ -0,0 +1,50 @@
+<p align="center">
+    <img src="./docs/assets/Bee_Dark.svg" height="128">
+    <h1 align="center">Bee API</h1>
+</p>
+OpenAI-compatible Assistants API backed by [Bee Agent Framework](https://github.com/i-am-bee/bee-agent-framework)
+
+## Getting started
+
+1. Create `.env` (from `.env.example`) and fill in values.
+2. Run `pnpm install` to install dependencies.
+3. Start the server with `pnpm start:dev`
+
+## Technologies
+
+- [Fastify](https://fastify.dev/) as the web framework
+- [MikroORM](https://mikro-orm.io/) backed by [MongoDB](https://www.mongodb.com/) as the database layer
+- [BullMQ](https://docs.bullmq.io/guide/jobs) backed by [Redis](https://redis.io/) as the job executor
+- [Bee Agent Framework](https://github.com/i-am-bee/bee-agent-framework) as the agent execution engine
+
+## Architecture overview
+
+The Assistants API consists mostly of CRUDL endpoints for managing API resources like assistants, threads, runs and more. Furthermore, some resources are asynchronous in a sense that they contain `status` changing over time as the background execution progresses. Clients use polling or streaming to watch for status updates of such resources.
+
+### Infrastructure
+
+The infrastructure consists of:
+- REST API server
+- MongoDB
+- Redis
+
+The REST API server stores resources in MongoDB database. Redis is used by BullMQ, rate limiter and as pub/sub broker for event streaming. Agent execution is performed by the Bee Agent Framework using various adapters for inference and embeddings.
+
+### Server
+
+The codebase contains several types of modules:
+-  `*.modules.ts` containing endpoint handlers
+-  `*.services.ts` containing services for the handlers
+-  `dtos/*.ts` containing JSON schema definitions for resources
+-  `*.entity.ts` containing ORM definitions for database entities
+-  `*.queue.ts` containing BullMQ queues and workers for asynchronous execution
+
+These modules are connected in the following manner
+
+```
+module ---> dto
+       ---> service ---> entity
+                    ---> queue ---> entity
+```
+
+OpenAPI schema is auto-generated from the `dtos` and exposed on the `/docs` endpoint.