rocketride-server 发布 vscode-v1.4.0：新增多类 Agent 工具节点与 LLM 适配层

Official release notes (https://github.com/rocketride-org/rocketride-server/releases/tag/vscode-v1.4.0) ### Added - **ai**: every subprocess exposes /task/data via a shared web server (#912) - **analytics**: shared, transport-agnostic telemetry core (loose report + app) (#1523) - **anonymize**: configurable entity types + token redaction style (#1447) - **chat-widget**: embeddable, brandable chat widget — web component + script-tag bubble (#1586) - **ci**: migrate Discord notifier workflows to forum channels (tags + auto-archive) (#1510) - **ci**: on-demand + monthly multi-OS compile/test matrix (#1537) - **ci**: keep Discord forum tags live (re-apply on state/label change via bot) (#1722) - **cli**: rocketride validate subcommand (Python + TS) + validate-pipes composite action (#1573) - **client-python**: `rocketride otel` — OpenTelemetry bridge over the ingester protocol (#1612) - **database**: Sequelize ORM over pipes + server-side DB transactions (#1467) - **deploy-2**: teams-as-environments deploy, owner-scoped task identity, push-driven surfaces (#1764) - **events-ui**: real-time DAP event monitor micro-frontend (#1484) - **explorer**: File Explorer app with rich media viewers, Monaco editor, hex viewer, and Open with… menu (#1356) - **llm**: virtualize provider interface behind a normalized Adapter (#1683) - **n8n-nodes**: add n8n-nodes-rocketride community node package (#1255) - **nodes**: add answer_documents node to bridge answers into documents (#1506) - **nodes**: add authoritative_overlay SEC period-scoped cross-check guard (#1430) - **nodes**: add chunker — sentence-boundary and tiktoken-based text chunking with configurable overlap and per-chunk character offsets; recursive character splitting stays with the General Text (`preprocessor_langchain`) node (#897) - **nodes**: add cognee node (#1501) - **nodes**: add context_optimizer node — fits questions into the model context window by token budget (#1489) - **nodes**: add currency_convert_explicit node (#1497) - **nodes**: add extract_facts node — document-context + cell-by-cell reader + validator (#1426) (#1545) - **nodes**: add n8n workflow-automation node (#1231) - **nodes**: add tool_oura Oura Ring connector (#1625) - **nodes**: add tool_slack — Slack agent-tool node (post messages, list channels, read history) (#1575) - **nodes**: tool_filesystem pipeline-sink lanes (#1651) - **nodes**: tool_google_workspace — shared Google client + gmail, sheets, docs, calendar, drive services (#1570) - **nodes**: add Guild.ai tool node (#1689) - **nodes**: add tool_pipedrive CRM tool node (#1687) - **nodes**: add db_firestore — Google Cloud Firestore Native-mode document get/set (#1511, #1515) - **nodes**: add tool_gcs — Google Cloud Storage list and download tool (#1512, #1515) - **nodes**: add tool_vertex_search — Vertex AI Vector Search nearest-neighbor tool (#1513, #1515) - **nodes**: add LaserData memory tool node (tool_laserdata_memory) (#1760) - **nodes**: add a FalkorDB URL connection profile to graph_falkordb (#1786) - **nodes**: add schema_validate node (#1520) - **nodes**: add normalize_facts node (#1518) - **nodes**: add tool_notion — Notion agent-tool node (search, read pages/database rows, create/update content) (#2141) - **nodes**: add search_hybrid — BM25 + vector RRF post-retrieval re-ranker (#899) - **nodes, ai**: add graph base class; make FalkorDB a real graph node (#1584) - **nodes, server**: migrate neo4j onto the graph base class (#1611) - **nodes,ai**: RocketRide cloud DB nodes — sql, vector, graph (#1713) - **nodes,vscode**: add Gmail tool node + Google user-OAuth broker (RR-1055, RR-1142) (#1334) - **run-logging**: task-event continuum, DVR sessions, permanent trace identity (#1661) - **server**: replace Breakpad with Crashpad crash reporting (#1476) (#1524) - **server, nodes**: add json lane (#1297) - **shared-ui**: pipeline TTL settings — toolbar cog + modal, VS Code support (RR-309) (#1521) - **status**: server-computed run report card, pipe-idle analytics, and DVR transport rework (#1678) - **storage**: joined filesystem, in-store authorization, task-file identity, and rocketlib.getTask (#1686) - **test-ui**: stress/chaos testing app for RocketRide backend (#1462) - **ui**: Tabulator DataGrid + record-panel standard, unified usePrefs, shell-api v1 freeze (#1632) - **ui**: UI consistency program: unify all apps on shared archetypes, components, and a versioned shell contract (#1540) - add HydraDB database node (db_hydradb) (#1499) - media stream descriptors + end-to-end source provenance (#1525) ### Fixed - **nodes**: harden HTTP Request SSRF guardrails for URL credentials and hostname-safe whitelist boundaries; regexes now match canonical URL starts and fail closed on unsupported or ambiguous authority syntax, so unrestricted authority and search-style patterns may require migration (#2060) - **agent**: optional require_tool_call runtime fabrication guard (#1490) - **ai**: install test deps via depends() so engine constraints apply (#1466) (#1471) - **ai**: migrate GPU import guard to find_spec/exec_module (Python 3.12+) (#1460) - **ai**: stop Surya `languages` from splitting model identity (#1749) - **ai**: keep GLiNER inference params out of model identity (#1750) - **ai**: send Whisper `language` per request instead of at load (#1751) - **auth**: register /auth/vscode/google outside standardEndpoints gate (#1543) - **auth**: log why has_permission denies instead of swallowing silently (#1670) - **build**: compare content, not mtime, when sync file sizes match (#1477) (#1567) - **build**: linux build improvement (#1473) - **build**: make Breakpad symbol generation non-fatal (#1595) (#1596) - **check-externals**: disable NLTK's CWE-427 import guard for the import checks (#1783) - **chroma**: validate server version and surface silent index failures (#1494) - **ci**: read full marker body in Discord issue/PR notifiers (#1530) - **client**: make connection lifecycle generation-safe (#1668) - **depends**: pin bootstrap tools and pydantic to stop downgrade churn (#1536) - **deps**: restore a resolvable onnxruntime pin — 1.20.1 is gone from PyPI (#1725) - **docs**: normalize Discord invite links to a single code (#1746) - **docs**: unblock docs deploys broken by tool_google_workspace's docs/ variant dir (#1757) - **endpoints**: unique stable per-pipe endpoint URLs ({host}/{type}/{project_id}/{source}) (#1557) - **explorer-ui**: correct hex viewer virtual scrolling for large files (#1633) - **llm_anthropic**: disable extended thinking at the node as an interim stop-gap (#1681) - **nodes**: accept documents input lane in extract_data (#1504) - **nodes**: friendlier chroma port + splitter-profile config (RR-1416) (#1496) - **nodes**: handle Windows MAX_PATH in Python output nodes (#1546) - **nodes**: reclassify FalkorDB as database, not store (classType) (#1580) - **nodes**: render Prompt node instructions as multi-line textarea (#1516) (#1564) - **nodes**: simplify n8n configuration UX (#1588) - **nodes**: warn that parse/llamaparse are not provenance-preserving (RR-1406) (#1507) - **nodes**: ocr tests (#1807) - **nodes**: gate currency_convert_explicit on isJson() (#1744) - **nodes, test**: scope core-module stubs so they don't leak across tests (#1642) - **ocr**: reader stringified the image bytearray instead of decoding it (#1796) - **ocr**: route node diagnostics to the engine log instead of discarding them (#1799) - **ocr**: documents lane called a non-callable reader and dropped its text (#1801) - **server**: bind Python instance.closing() to cb_closing (#1667) - **server**: dispatch pipeline lifecycle in topological order, including control sub-pipelines (#1669) - **server**: simplify pipeline config validation (#1791) - **shared-ui**: persist Google OAuth tokens to node config and stop authType flip (#1548) - **shared-ui,dropper-ui**: render media lanes as players instead of dumping base64 into the trace webview (#1619) - **shared-ui,vscode**: key OAuth tier scopes by node provider for least privilege (#1579) - **shell-ui**: define REACT_APP_OAUTH_ROOT_URL (ReferenceError: process is not defined) (#1486) - **shell-ui**: route all auth entries to the Zitadel login page (register via its built-in link) (#1556) - **shell-ui**: surface dashboard fetch failures instead of loading forever (#1673) - **shell-ui**: distinguish network from auth failures and surface both in a recovery banner (#1553) - **storage**: version the /task/fetch claim so pre-upgrade tokens cannot be reinterpreted (#1770) - **tool_mcp_client**: normalize zero-argument tool schemas for reasoning-model agents (#1491) - **vscode**: auto-populate .env with resolved engine URI/key on connect (#1492) - **vscode**: group Deep Agent Subagent under AGENT in node picker (#1505) (#1517) - **vscode**: repoint dead example links to awesome-rocketride (#1422) (#1448) - **vscode**: wire native file dialog for embedded-app Browse button (#1011) (#1235) - **weaviate**: pin grpcio-health-checking= 3.19 version gate - **builder** — Prevent deadlock when conditional skips a dedup branch - **canvas** — Forward drag events through CreateNodePanel backdrop - **canvas** — Improve QuickAdd popup label copy - **canvas** — Pipeline render-loop fix + sidebar logo home nav - **checkout** — Carry validated promo into preselected-plan checkout (#1604) - **checkout** — Open plan CTA links externally in the VS Code extension - **chroma** — Exclude soft-deleted documents from default searches - **ci** — Add build passthrough for docs-only PRs - **ci** — Add id-token: write to nightly docker job to unblock prereleases - **ci** — Add id-token: write to release docker job (preemptive fix) - **ci** — Allow Nightly workflow_dispatch from stage and main - **ci** — Block shell injection in _build.yaml via env-var indirection - **ci** — Drop artifact retention to 1 day to fit in 0.5 GB storage cap - **ci** — Grant nightly workflow workflows:write to publish prereleases - **ci** — Match GraphQL github-actions author login in marker search - **ci** — Pin all actions to immutable commit hash - **ci** — Pin cryptography at build + UV_LINK_MODE=copy to stop Windows nodes:test _rust.pyd lock - **ci** — Prevent duplicate Discord posts on discussion edits - **ci** — Push release/prerelease tags via WORKFLOW_PAT (repo+workflow scope) - **ci** — Scope nightly.yaml writes to per-job permissions - **ci** — Stop Discord discussions notifier from posting blank bot comments - **ci** — Stop Discord notifier from posting blank bot comments - **ci** — Use literal ROCKETRIDE_APIKEY in Test step instead of secret - **ci/cd, nodes** — Full tests - **cli** — Catch ValueError from json5.loads, not non-existent json5.JSONError (#1379) - **cli** — Implement signal handlers for graceful shutdown (RR-655) - **client-python** — Bound truncate_filename for small widths (#1385) - **client-python** — Break drain cycle causing RecursionError on disconnect - **client-python** — Clean pending DAP request on send failure - **client-python** — Remove token='*' from restart command - **client-ts** — Preserve user-specified port in normalizeUri - **clients** — Pipe errors (Python + TS), SSE rollback, getTaskStatus timeout - **connection** — Handle missing hostname in URI parsing - **core** — _message_tasks set is accessed from multiple async contexts without a lock - **crewai** — Patch async compatibility issues for CrewAI 1.14.x hierarchical crews - **dap** — Don't let an oversized subprocess stdout line kill the pipeline task - **database** — Bound Neo4j EXECUTE rows, parse allow_execute strictly - **database** — Gate EXECUTE path, bound results, validate SDK inputs - **deepagent** — Avoid transformers dependency for token counting - **depends** — Bootstrap setuptools alongside wheel - **depends,whisper** — Unicode encoding + whisper GPU probe in server mode - **deps** — Restore libc fields in pnpm-lock.yaml - **deps** — Update pnpm-lock.yaml for lucide-react dependency - **discord-discussions** — Normalize webhook base, at-most-once POST - **docker** — Provision /opt/data and use /version for healthcheck - **docs** — Documenation update - **docs** — Trigger workflow on README.md changes to keep llms.txt current - **docs** — Update source connector docs - **easyocr** — Unwrap DataParallel to prevent SIGABRT under concurrent load - **engine** — Forward --node_path to task subprocess so workspace-local nodes load - **engine** — In-process clients connect to wrong port when engine binds to OS-assigned port - **engine** — Pass uv --excludes relative to cwd so spaced install paths resolve - **engine** — Pass uv -c constraints relative to cwd so spaced install paths resolve - **engine** — Progress sidecar for depends.py, remove dry-run, exclude crewai-cli - **git** — Lefthook sequential - **guardrails** — Reject '|' in email TLD char class - **home-ui** — SaaS pipeline export, host-slot sidebar refactor & shell/canvas polish - **icons** — Add auto-currentcolor theming and fix node icon assets - **icons** — Preserve original colors for Baidu Qianfan node icon - **llm** — Clear correct _chat reference in endGlobal across all LLM nodes - **llm_anthropic** — Drop temperature, removed on Opus 4.7+ - **llm_ollama** — Auto-configure reasoning models (temperature + reasoning_effort) - **milvus** — Apply retrieval_score_threshold filter in searchSemantic (#642) - **mixins** — Error handling for missing pipeline configuration files - **nodes** — Add input validation/sanitization to LLM chat drivers - **nodes** — Add rate limiting to tool_http_request node - **nodes** — Agent keyword search returns nothing (RR-1363) - **nodes** — Bootstrap deps in agent_rocketride and drop transformers from llm_anthropic - **nodes** — Correct GMI Cloud Gemini model IDs (#996) - **nodes** — Enforce LF for .md files and harden doc generator on Windows - **nodes** — Improve error handling in remote node execution - **nodes** — Openai - add latest openai models - **nodes** — Reclassify FalkorDB as database, not store (classType) [stage] - **nodes** — Remove caption + face_detection nodes ahead of release - **nodes** — Remove deprecated gemini models - **nodes** — Remove llm_vertex, fix llm_gemini key handling, migrate vision nodes to LLMBase - **nodes** — Remove redundant full-res image encode/decode round-trips - **nodes** — User-configured interval ignored in frame grabber - **ocr** — Support img2table 2.0 OCRInstance API rewrite - **profiler-ui** — Rehydrate report on mount so split panes keep the chart - **qdrant** — Correct score-space mismatch and isDeleted filter - **release** — Drop orphaned TWINE_USERNAME/PASSWORD lines inside MCP PyPI run block - **release** — Durable CHANGELOG-sourced GitHub Release notes - **rocketlib** — Satisfy ruff D204/Q000 in __init__.pyi stub - **scripts** — Guard state.js dot-paths against prototype pollution - **scripts** — Resolve undefined BUILD_ROOT and DIST_DIR in licenses.js - **scripts** — Skip pytest tasks when target dir has no test files - **sdk** — Align CreditBalance type with backend camelCase wire shape - **security** — Expand sensitive field redaction patterns in VS Code logger - **security** — Explicit branch check on nightly workflow_run trigger - **security** — Prevent arbitrary module injection via /use endpoint (RCE) - **security** — Prevent filter expression injection in Milvus vector store - **security** — Require authentication for profiler endpoints and escape HTML output - **security** — Restrict environment variable expansion in pipeline configs - **security** — Switch SQL safety check from blacklist to whitelist approach - **security** — Use pathlib. Repository README (current repository documentation, not release changes) RocketRide is the open source AIDE: the AI Development Environment. Build, deploy and harness production-ready AI solutions at light speed, all within your IDE or using the CLI in your terminal.       Your code editor just became the AIDE. RocketRide turns the classic IDE you already use into a full AI Development Environment: one place to compose, debug, observe, and deploy AI runtimes using any model, any tool, any framework, with zero vendor lock-in. Equipped with deep observability and backed by a battle-tested, high-throughput C++ engine, what you build is production-ready the moment it runs. It's the harness for everything behind your AI applications, not just the agents, but the whole stack beneath them. Under the hood, RocketRide is an open source data pipeline builder and runtime built for AI and ML workloads. With 100+ pipeline nodes spanning 15+ LLM providers, 9 vector databases, OCR, NER, and more, pipelines are defined as portable JSON, built visually in VS Code, and executed by a multithreaded C++ runtime. From real-time data processing to multimodal AI search, RocketRide runs entirely on your own infrastructure. Home | Documentation | Python SDK | TypeScript SDK | MCP Server | VS Code Marketplace | Open VSX Registry _Design, test, and ship complex AI workflows from a visual canvas, right where you write code._ _Drop pipelines into any Python or TypeScript app with a few lines of code, no infrastructure glue required._ ## Where to start Run it locally Install the extension, pick Local, and run a pipeline on your own machine. Nothing to sign up for. Quick Start → Shipping your first AI feature? Start from what you already know: a working pipeline in three steps, no API keys. Start here → Want to contribute? Claim an issue with one comment, add a node in Python, or fix the docs you're reading. Contributing → Want to run it without ops? We operate the engine; your .pipe runs unchanged, with nothing to provision. RocketRide Cloud → **Open source, MIT.** The whole engine is MIT-licensed and OSI-compliant. No enterprise edition, nothing behind a paywall. ## Features | Feature | Description | | :-------------------------------- | :----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | **Visual Pipeline Builder** | Drag, connect, and configure nodes in VS Code, no boilerplate. Real-time observability tracks token usage, LLM calls, latency, and execution. Pipelines are portable JSON, version-controllable, shareable, and runnable anywhere. | | **High-Performance C++ Runtime** | Native multithreading purpose-built for the throughput demands of AI and data workloads. No bottlenecks, no compromises for production scale. | | **100+ Pipeline Nodes** | 15+ LLM providers, 9 vector databases, OCR, NER, PII anonymization, chunking strategies, embedding models, and more. All nodes are Python-extensible, build and publish your own. | | **Multi-Agent Workflows** | Built-in CrewAI and LangChain support. Chain agents, share memory across pipeline runs, and manage multi-step reasoning at scale. | | **Coding Agent Ready** | RocketRide auto-detects your coding agent: Claude, Cursor, and more. Build, modify, and deploy pipelines through natural language. | | **TypeScript, Python & MCP SDKs** | Integrate pipelines into native apps, expose them as callable tools for AI assistants, or build programmatic workflows into your existing codebase. | | **Zero Dependency Headaches** | Python environments, C++ toolchains, Java/Tika, and all node dependencies managed automatically. Clone, build, run, no manual setup. | | **One-Click Deploy** | Run on Docker, on-prem, or [RocketRide Cloud](https://cloud.rocketride.ai/). Production-ready architecture from day one, not retrofitted from a demo. | ## Quick Start 1. Install the extension for your IDE. Search for RocketRide in the extension marketplace: [Not seeing your IDE? Open an issue](https://github.com/rocketride-org/rocketride-server/issues/new) · [Download directly](https://open-vsx.org/extension/RocketRide/rocketride) 2. Click the RocketRide extension in your IDE 3. Deploy a server - you'll be prompted on how you want to run the server. Choose the option that fits your setup: - **Local (Recommended)** - This pulls the server directly into your IDE without any additional setup. - **On-Premises** - Run the server on your own hardware for full control and data residency. Pull the image and deploy to Docker or clone this repo and [build from source](CONTRIBUTING.md#getting-started). ## Your first AI feature _Whether you build web apps, integrate APIs, or run backend services, you already have the mental model._ | You already know | Same idea in RocketRide | | :-- | :-- | | A route that receives a request | A **source node**: `webhook`, `chat`, or `dropper` (file drop) | | Middleware chained in order | **Nodes**, wired together on a canvas. Most are Python you can open and read | | The response you return | A **response** node | | Config in git, like a Dockerfile | The `.pipe` file: plain JSON, diffable, reviewable | | Calling a service from your app | The Python or TypeScript SDK: one call in, one result out | **Three steps, no API keys:** 1. **Run it locally.** Follow the [Quick Start](#quick-start) and pick **Local** when asked. That's the whole install. 2. **Open a working pipeline.** Open [`examples/document-processor.pipe`](examples/document-processor.pipe) in your IDE. Give it a PDF or an image and it pulls out the text (including text inside images), finds names, addresses and other personal data, and returns a cleaned copy. It runs on local models, so nothing leaves your machine. Press ▶ on the source node to start it. Its source is a `webhook`, so it waits for input — step 3 is how you send some. 3. **Call it from your code.** The pipeline is now a function your service can call. Run it from the folder you opened in your IDE: the extension writes the engine's connection details into a `.env` there, and the SDK reads them automatically. Run it from anywhere else and the client falls back to RocketRide Cloud instead. ```bash pip install rocketride ``` ```python import asyncio from rocketride import RocketRideClient async def main(): async with RocketRideClient() as client: run = await client.use(filepath='examples/document-processor.pipe') out = await client.send( run['token'], 'Alice Smith, 12 Elm St, Springfield.', objinfo={'name': 'note.txt'}, mimetype='text/plain' ) print(out) await client.terminate(run['token']) asyncio.run(main()) ``` TypeScript works the same way: [`npm install rocketride`](https://www.npmjs.com/package/rocketride) · [SDK docs](https://docs.rocketride.org/clients/typescript) **Next steps, one at a time:** - **Add a model.** Drop an `llm_*` node between the source and the response and add one API key. Same pipeline, now with an LLM in the loop. - **Search by meaning.** [`examples/rag-pipeline.pipe`](examples/rag-pipeline.pipe) is the standard "ask questions about my documents" pattern. Needs one LLM key and a vector database ([Qdrant](https://qdrant.tech/) can run locally). - **Give it tools.** An **agent** node is an LLM allowed to call other nodes in a loop until it's done. [`examples/agent-workflow.pipe`](examples/agent-workflow.pipe) shows the shape. Words you'll see, translated | Term | What it means here | | :-- | :-- | | **Pipeline** | A request handler built from steps, saved as a `.pipe` JSON file | | **Node** | One step. Providers (OpenAI, Anthropic…), tools (GitHub, Slack…), stores, parsers | | **Embedding** | Turning text into a list of numbers so "similar meaning" becomes "nearby numbers" | | **Vector database** | A store that finds records by meaning instead of exact match | | **RAG** | Retrieve the relevant documents first, then ask the model with them in context | | **Agent** | An LLM that can call tools and nodes repeatedly until the task is done | | **OCR / NER / PII** | Read text out of images / find names, dates, organisations / detect personal data | | **Chunking** | Splitting long documents into pieces small enough to embed and retrieve | ## Building Your First Pipe 1. All pipelines are recognized with the `*.pipe` format. Each pipeline and its configuration are JSON objects - but the extension in your IDE will render within our visual builder canvas. 2. All pipelines begin with a source node: _webhook_, _chat_, or _dropper_. For specific usage, examples, and inspiration on how to build pipelines, check out our [guides and documentation](https://docs.rocketride.org/). 3. Connect input lanes and output lanes by type to properly wire your pipeline. Some nodes like agents or LLMs can be invoked as tools for use by a parent node as shown below: 4. You can run a pipeline from the canvas by pressing the ▶ button on the source node or from the `Connection Manager` directly. 5. Deploy your pipelines - pick the path that fits: - **Docker** - Download the RocketRide server image and create a container. Requires [Docker](https://docs.docker.com/get-docker/) to be installed. ```bash docker pull ghcr.io/rocketride-org/rocketride-engine:latest docker create --name rocketride-engine -p 5565:5565 ghcr.io/rocketride-org/rocketride-engine:latest ``` - **Local Deployment** - Download your preferred runtime as a standalone process from the **Deploy** page in the `Connection Manager`. - **RocketRide Cloud** - Skip the setup and ship straight to [managed hosting](https://cloud.rocketride.ai/). Same portable pipeline JSON, zero infrastructure to run, from prototype to production. [Get started](https://cloud.rocketride.ai/) 6. Run your pipelines as standalone processes or integrate them into your existing [Python](https://docs.rocketride.org/clients/python) and [TypeScript/JS](https://docs.rocketride.org/clients/typescript) applications utilizing our SDK. ## Observability Selecting running pipelines allows for in-depth analytics. Trace call trees, token usage, memory consumption, and more to optimize your pipelines before scaling and deploying. Find the models, agents, and tools best fit for your task. ## Documentation Everything on [docs.rocketride.org](https://docs.rocketride.org/) is built from the [`docs/`](docs/) tree in this repo, split by audience. [`docs/README.md`](docs/README.md) is the full map; the short version: | Looking for | Start at | |---|---| | Using RocketRide: quickstart, concepts, guides, cloud, self-hosting | [`docs/public/product/`](docs/public/product/) (the site, page for page) | | SDK guides | [TypeScript](docs/public/typescript/) · [Python](docs/public/python/) · [MCP](docs/public/mcp/) | | Building your own client on the engine protocol | [`docs/public/product/connect/websocket/`](docs/public/product/connect/websocket/) | | Contributing: setup, builder, engine internals, node authoring | [`docs/development/`](docs/development/index.md) | | Node catalog | each node's `README.md` under [`nodes/src/nodes/`](nodes/src/nodes) | | Apps: VS Code extension and the shell UIs | each app's own folder under [`apps/`](apps/); how they fit the monorepo in [`docs/development/apps/`](docs/development/apps/index.md) | | Pointing an AI assistant at RocketRide | [`docs/agents/`](docs/agents/README.md) | Repo-wide contributor rules live in [AGENTS.md](AGENTS.md). ## Contributing **Good places to start** - **Claim an issue with one comment.** Comment `/assign` on any open issue and it's yours. No permissions or membership needed. - **Good first issues** are labelled [`good first issue`](https://github.com/rocketride-org/rocketride-server/issues?q=is%3Aissue+is%3Aopen+label%3A%22good+first+issue%22); `help wanted` marks bigger ones we'd like a hand with. - **Add a node.** Every node is a small Python package under [`nodes/src/nodes/`](nodes/src/nodes). A new provider, tool, or store makes a good first contribution. Guide: [Adding a New Node](docs/development/nodes/index.md#adding-a-new-node). - **Fix what you just read.** Docs PRs are welcome, including this README. **How it works** - Fork, branch as `/RR--`, open a PR against `develop`, link the issue. Full process, style guides, and test commands: [CONTRIBUTING.md](CONTRIBUTING.md). - One build tool for the whole monorepo: `./builder` manages the C++ toolchain, Python environments, and Java/Tika for you. Clone, build, run. - Roles and decisions: [GOVERNANCE.md](GOVERNANCE.md) · Releases: [RELEASE.md](RELEASE.md) · Security reports: [SECURITY.md](SECURITY.md) · Questions: [Discussions](https://github.com/rocketride-org/rocketride-server/discussions), [Discord](https://discord.gg/PMXrtenMsY), [SUPPORT.md](.github/SUPPORT.md) ## Two ways to run RocketRide _Let us handle the infrastructure, or own every layer._ ROCKETRIDE CLOUD  ·  NOW LIVE ON-PREM Let us run it, zero ops The fastest way to get started. We operate the engine; you point a client at the endpoint and build. The same .pipe file runs unchanged, with no infrastructure to provision, from prototype to production. Run it yourself, free Docker, on-prem, or local. Full control and data residency. Open source, MIT, no lock-in, ever. Get Started Quick Start ## RocketRide Cloud  NOW LIVE ### Run pipelines, not infrastructure. With RocketRide Cloud, offload the complex, heavy lifting of production AI. Build your pipeline once, in the same portable .pipe format, and we run it, scale it, and keep it fast. No servers, no ops, no rewrite from prototype to production. A fraction of the cost Our model server runs your AI workloads far more efficiently, so you pay a fraction of what standard hosting costs. Collaborate as a team Work together on shared pipelines. Your whole team builds, runs, and iterates in one place. No-hassle infra We handle everything: servers, scaling, upgrades, uptime. Nothing to provision, nothing to operate. Higher performance, built-in scale More throughput on the multithreaded C++ engine, with automatic scalability built in. Connecting takes two lines. Same portable pipeline JSON, now hosted for you: ```bash ROCKETRIDE_URI=https://api.rocketride.ai ROCKETRIDE_AUTH=your-api-token ``` Get Started   ·   Read the Cloud docs ## On-Prem  FREE & MIT ### Own every layer. Run the very same engine yourself, wherever your data lives. Docker, on-premises, or a local process in your IDE. Full control, full data residency, and zero lock-in. The exact .pipe you build on Cloud runs unchanged on your own hardware. Full control and data residency Your data and model calls never leave your infrastructure. Run it behind your firewall, or fully air-gapped. Open source, MIT The whole engine is MIT-licensed and OSI-compliant. No enterprise edition, nothing behind a paywall. Runs anywhere Docker, on-premises, bare metal, or local. Scale out to a cluster with the Helm chart when you need to. The same C++ engine Identical multithreaded runtime and execution semantics as Cloud. Move a pipeline between them anytime. Point a client at your local engine in one line: ```bash ROCKETRIDE_URI=ws://localhost:5565 ``` Quick Start   ·   Read the On-Prem docs Ready to ship? Deploy your pipeline on RocketRide Cloud, or run on-prem, free. ## Contributors RocketRide is built by a growing community of contributors. Whether you've fixed a bug, added a node, improved docs, or helped someone on Discord, thank you. New contributions are always welcome - check out our [contributing guide](CONTRIBUTING.md) to get started. --- Made with ♥ in SF & EU