QTSurfer beta
Documentation

Open source

Index of the open-source projects QTSurfer publishes on GitHub — API clients and SDKs, the MCP server and strategy skills, the Lastra data format and its tooling, market-data streaming, a charting component, and the API contract and documentation.

QTSurfer publishes the libraries, formats, and tools around the platform as open source under the QTSurfer organisation on GitHub. This page indexes them: what each project is for, where to find it, and which page of these docs covers it in more depth. Code is published under the Apache-2.0 licence; the documentation prose in qtsurfer-content is CC BY 4.0. Each repository’s README states its own terms, and is the authority for them.

Clients and SDKs

Talk to the QTSurfer API from your own code. The SDKs wrap the API in complete workflows; the API clients are generated from the OpenAPI contract and map one to one onto its endpoints. See Clients and SDKs for how to choose between them.

  • sdk-java — the Java SDK. Authentication, the compile, prepare, and execute workflow with retries and cancellation, live execution over WebSocket, and normalised errors. Published through JitPack.
  • sdk-ts — the TypeScript SDK, with the same workflow orchestration, a single-call backtest flow, and live execution. On npm as @qtsurfer/sdk.
  • sdk-python — the Python SDK, with workflow orchestration, token refresh on expiry, and a pluggable token store. On PyPI as qtsurfer-sdk.
  • api-client-java — the generated Java client for the REST API. Published through JitPack.
  • api-client-ts — the generated TypeScript client: fully typed, tree-shakeable, built on the native fetch. On npm as @qtsurfer/api-client.
  • api-client-python — the generated Python client. On PyPI as qtsurfer-api-client.

AI assistants

  • mcp-java — the QTSurfer Model Context Protocol server. It exposes account, live execution, backtesting, and market data as tools to any MCP-capable assistant over standard input and output. Installation and the tool list are in MCP server.
  • strategy-skills — agent skills for writing, reviewing, and debugging QTSurfer strategies in Java and in QTScript. They follow the open Agent Skills format, so they work with Claude Code, Codex, Cursor, Cline, and other agents that support it. The strategy guides in these docs are generated from them.

Lastra and data formats

The hourly market-data segments QTSurfer serves are files you can keep, query, and convert. The Lastra format, its readers and writers, and the converters are open. See Lastra for the shortest path from a downloaded segment to a DataFrame, a SQL query, or a Parquet file.

  • lastra-java — the reference implementation of Lastra, a columnar file format for numeric time series with a codec per column, row groups, and CRC32 integrity checks. It also holds the specification of the wire format. Published through JitPack.
  • lastra-py — a Python reader and writer, bit-exact with the Java implementation. On PyPI as lastra.
  • lastra-ts — a TypeScript reader for the browser and Node.js, with Apache Arrow interoperability. On npm as @qtsurfer/lastra.
  • duckdb-lastra — a DuckDB extension that queries .lastra files as tables with timestamp pushdown, including remote files over HTTP range requests.
  • lastra-convert — a command-line converter, with a Java API, between Lastra, Parquet, and CSV.
  • lastra-convert-py — the Python counterpart, which also converts Apache Arrow. On PyPI as lastra-convert.
  • alp-java — a dependency-free Java implementation of ALP, the adaptive lossless floating-point compression that Lastra uses for decimal columns such as prices. It follows the published algorithm of Afroozeh and Boncz (SIGMOD 2024).
  • alp-py — the Python implementation of ALP, byte-compatible with alp-java. On PyPI as alp-codec.
  • parquet-lite — a Java library to read and write Apache Parquet files without the Hadoop dependency tree. It is a fork of strategicblue/parquet-floor. Published through JitPack.

Market-data streaming

  • qtstreamx — normalised market-data streaming for the JVM. Exchange WebSocket feeds and on-chain Uniswap swaps sit behind one set of interfaces and one record model (tickers, klines, funding rates, trades), and the codec, WebSocket client, and transport are separate modules you can swap.

Visualisation

  • svelte-timeseries — a Svelte component for exploring very large time series in the browser. It reads Parquet, Arrow, and Lastra through DuckDB-WASM, and renders with either Apache ECharts or TradingView Lightweight Charts, with markers and events overlaid. On npm as @qtsurfer/svelte-timeseries, with a live demo.

API contract and documentation

  • qtsurfer-api — the OpenAPI specification, the AsyncAPI contract of the live-execution WebSocket, and the Markdown guides behind the API section of these docs. The interactive reference is published at qtsurfer.github.io.
  • qtsurfer-engine-java-docs — the published Java API reference for the types a strategy author uses: strategy base classes, the market-data model, the indicator catalogue, and the fee, slippage, and fill models. It is browsable on this site as the Java API reference.
  • qtsurfer-content — the source of the documentation, Learn articles, and glossary on this site, in every published language. It accepts corrections as pull requests.