Aerospace MCP

# Product Requirements: Advanced Aerospace MCP Tools ## Overview Expand this MCP server beyond flight planning to expose research-grade tools across aircraft, drones, rockets (guided/unguided), and spacecraft. Favor lightweight, reproducible calculations with optional heavy integrations. All tools return structured, unit-consistent results and graceful fallbacks when an optional dependency is missing. ## Goals - Provide domain-focused MCP tools with clear inputs/outputs and SI units. - Wrap widely used Python libraries to accelerate aerospace research workflows. - Keep core server responsive; long jobs run with timeouts and clear errors. ## Non-Goals - Real-time flight operations, ATC integration, or safety-critical usage. - Full CFD/FEA pipelines; provide lower-order aero methods and hooks instead. ## Tool Catalog (Proposed) Notes: Tools are grouped; each lists candidate libraries. Prefer Tier 1 (light) first. 1) Common Atmosphere & Frames - get_atmosphere_profile: ISA/COESA properties vs altitude, lapse-rate banding. Libraries: ambiance (Tier 1), aerosandbox.atmosphere (Tier 1) - wind_model_simple: logarithmic or power-law profile for low-altitude studies. Libraries: none (in-house), optional: windscrape+xarray (Tier 3) - transform_frames: ECI/ECEF/ITRF conversions, timescales. Libraries: astropy (Tier 1), skyfield (Tier 1), spiceypy (Tier 2) 2) Aircraft Aerodynamics & Propulsion - wing_vlm: VLM-based CL/CD/CM, stability derivatives. Libraries: AeroSandbox (Tier 1), MachUpX (Tier 1) - airfoil_polar: Airfoil polars via XFoil wrapper or surrogate. Libraries: aerosandbox (built-in XFoil), xfoil-python (Tier 2) - turbofan_cycle: On-design/off-design thrust and SFC maps. Libraries: pyCycle + OpenMDAO (Tier 2) - mission_performance: Mission fuel/time using Breguet + OpenAP. Libraries: OpenAP (existing), aerosandbox (mission models) 3) Drone/Propeller Systems - propeller_bemt: Thrust/torque/eta vs RPM, diameter, pitch. Libraries: AeroSandbox (BEMT), pybemt (Tier 2) - uav_energy_estimate: Multirotor endurance/range vs mass, battery, wind. Libraries: in-house + aerosandbox utilities - motor_match: Motor/ESC/prop selection sweep. Libraries: in-house; data-driven configs 4) Rockets (Unguided/Guided) - rocket_mass_budget: Stage mass fractions and ∆v (Tsiolkovsky). Libraries: in-house; RocketCEA for Isp (Tier 2) - rocket_vertical_ascent_3dof: Ballistic/lofted ascent with drag and winds. Libraries: RocketPy (Tier 1) - launch_trajectory_opt: Pitch program optimization with path constraints. Libraries: Dymos + OpenMDAO (Tier 2), casadi (Tier 2) 5) Spacecraft & Orbital Design - lambert_solve: Two-body Lambert solutions and transfer elements. Libraries: poliastro (Tier 1), pykep (Tier 2) - hohmann_bielliptic: ∆v budgets and timing. Libraries: poliastro (Tier 1) - orbit_propagate: Keplerian + J2 propagation, ephemerides. Libraries: poliastro+astropy (Tier 1), orekit (Tier 2), spiceypy (Tier 2) - porkchop_grid: Departure/arrival grid with C3/∆v contours. Libraries: poliastro (Tier 1) 6) GNC & Controls - kalman_filter_sim: IMU/GNSS fusion (EKF/UKF) on sample trajectory. Libraries: filterpy (Tier 1) - control_design: LQR/PID synthesis for linearized models. Libraries: python-control (Tier 1), slycot optional ## API Shape (MCP) - Each tool: name, JSON schema input, deterministic text/JSON output. - Units: SI by default; accept/return unit tags when practical. - Capability discovery: get_system_status lists available domains and missing deps. - Timeouts: default 10–60s; return partials or errors with actionable hints. ## Library Integration Plan - Add optional extras to `pyproject.toml`: - aircraft: aerosandbox, machupx - rockets: rocketpy, rocketcea - space: poliastro, astropy, spiceypy - trajopt: openmdao, dymos, casadi - gnc: filterpy, control - frames: skyfield - Lazy import inside wrappers; set AVAILABLE flags per feature. - For heavy stacks (OpenMDAO/Dymos, Orekit), mark Tier 2 and degrade gracefully. ## Implementation Plan (for AI Coding Tool) 1) Scaffolding - Create `aerospace_mcp/integrations/` with modules: - atmosphere.py, frames.py, aero.py, propellers.py, rockets.py, orbits.py, gnc.py, trajopt.py - Add `__all__` and small pydantic models for IO data structures. 2) Optional Dependencies - Update `pyproject.toml` with `[project.optional-dependencies]` groups listed above. - Add import probes: e.g., `AEROSANDBOX_AVAILABLE`, `POLIASTRO_AVAILABLE`. - Extend `get_system_status` to report availability counts/versions. 3) Wrappers (Tier 1 first) - atmosphere.get_atmosphere_profile(altitudes_m: list[float]) -> list[dict] - frames.transform_frames(epoch_iso: str, xyz: list[float], from: str, to: str) -> dict - aero.wing_vlm(planform: dict, alpha_deg: list[float], mach: float) -> dict - propellers.propeller_bemt(params: dict, rpm: list[float]) -> dict - rockets.rocket_mass_budget(stages: list[dict]) -> dict - rockets.vertical_ascent_3dof(vehicle: dict, env: dict) -> dict (RocketPy) - orbits.lambert_solve(r1_km: list[float], r2_km: list[float], tof_s: float, mu_km3s2: float) -> dict - orbits.hohmann_bielliptic(elements: dict) -> dict - orbits.orbit_propagate(elements: dict, tspan_s: float) -> list[dict] - gnc.kalman_filter_sim(traj_true: list[dict], sensors: dict) -> dict 4) MCP Tool Definitions - Add new Tool entries in `aerospace_mcp/server.py` with JSON Schemas mirroring the wrappers. - Standardize responses: headline, key scalars, and optional tabular blocks. - Respect timeouts with `asyncio.wait_for`; catch ImportError and return guidance. 5) Tests - Add unit tests per wrapper using synthetic inputs and golden values (small tolerances). - Skip tests when dependency unavailable via `pytest.importorskip`. - Add integration tests for representative tools (e.g., lambert_solve, propeller_bemt). 6) Docs & Examples - Update README “Features” to include new domains. - Add docs pages under `docs/` with usage examples and input templates. - Provide quick recipes (e.g., LEO transfer, 3DOF rocket ascent, UAV endurance). 7) Quality & Safety - Enforce SI units and document conversions; include NM/KM helpers already present. - Input validation via pydantic; clamp unphysical values with warnings. - Deterministic seeds where randomized. - Performance: add `functools.lru_cache` for atmospheric/ephemeris helpers. ## Acceptance Criteria - Tier 1 tools implemented and discoverable via `get_system_status`. - Meaningful outputs for at least: - Atmosphere profile - Wing VLM sweep or Airfoil polar - Propeller BEMT sweep - Rocket mass budget - 3DOF rocket ascent (basic) - Lambert transfer + Hohmann - Simple EKF demo - Tests pass in a “core” environment (no Tier 2 deps) with coverage on wrappers. ## Risks & Mitigations - Heavy deps (OpenMDAO/Orekit) complicate installs → keep optional, provide stubs. - Numerical instability (optimizers) → constrain defaults, document solver options. - Ephemeris data requirements → clearly document SPICE kernels; default to simplified models. ## Milestones - M1 (Core, 1–2 weeks): Atmosphere, Lambert/Hohmann, Rocket budget, Propeller BEMT. - M2 (Aircraft/Drone, 2 weeks): VLM wing, UAV energy, EKF. - M3 (Rockets/TrajOpt, 3–4 weeks): 3DOF ascent, basic pitch optimization. - M4 (Space/Advanced, 3–4 weeks): Porkchop, J2 propagation, SPICE/OREKIT optional.