openral-reasoner

OpenRAL S2 reasoner — the event-driven slow planning loop (CLAUDE.md §3 Layer 4). ReasonerCore consumes a WorldState snapshot plus rolling event buffers (failures, perception events, operator prompts) and emits exactly one typed ReasonerToolCall per tick via the LLM's structured tool-use API — no free-form JSON. The ROS-side reasoner_node (in packages/openral_reasoner_ros/) wraps this core with rclpy subscriptions and dispatch plumbing.

Layer

CLAUDE.md §3 Layer 4 — S2 slow reasoning: event-driven with a ~0.2 Hz heartbeat, sitting between the WorldStateAggregator and the S1 skill executor. Dispatch is direct typed tool calls.

Design

  • ROS 2 reasoner supervisor: direct typed tool-call dispatch.
  • Symbolic S2 reasoner: authored playbooks, self-maintained memory, success-gated task queue.
  • VLM-adjudicated completion, grounding-before-decompose, detection identity.

Public surface

from openral_reasoner import (
    ReasonerCore, ReasonerTickResult,          # the S2 tick loop
    ToolPalette, build_tool_palette,           # registry -> LLM tool palette
    AnthropicToolUseClient,                    # provider clients
    OpenAICompatibleToolUseClient,
    build_tool_use_client_from_env,            # model-first env selection
    ContextRenderer,                           # WorldState/event -> prompt context
    MemoryStore, MissionState,                 # self-maintained memory + mission ladder
    CriticWatchdog, SpatialMemoryQuerier,      # critic gating + spatial recall
    plan_active_search,                        # active-search frontier planning
)
  • ReasonerCore — the tick loop: render context → call the LLM with the tool palette → validate into a ReasonerToolCall → bounded replanning ladder on failure (retry → param-tweak → substitute-skill → goal-replan → human-handoff).
  • AnthropicToolUseClient / OpenAICompatibleToolUseClient — concrete ToolUseClient implementations; selected at activate-time from the curated openral_core.REASONER_MODELS registry via OPENRAL_REASONER_MODEL. OPENRAL_REASONER_ENDPOINT overrides where the model runs. No hidden default.
  • ToolPalette / build_tool_palette — generated from the local skill registry, rebuilt on /openral/skill_registry_changed.
  • MemoryStore, MissionState, evaluate_task_verdict — self-maintained memory and success-gated task queue.

See docs/methods/06-reasoning-wam-safety-observability.md for the full symbol inventory and docs/reference/reasoner-design.md for the design walkthrough.

Key dates

PyPI data
Data sourced directly from PyPI's database.
  • Released:
Newer release available (0.3.1)

1 maintainer

PyPI data
Data sourced directly from PyPI's database.
Avatar for AdrianLlopart from gravatar.com AdrianLlopart

License

Apache-2.0

Requires

Python <3.13, >=3.12

Report project as malware

Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

openral_reasoner-0.2.0.tar.gz (95.8 kB view details)

Uploaded Source

Built Distribution

If you're not sure about the file name format, learn more about wheel file names.

openral_reasoner-0.2.0-py3-none-any.whl (100.1 kB view details)

Uploaded Python 3

File details

Details for the file openral_reasoner-0.2.0.tar.gz.

File metadata

  • Download URL: openral_reasoner-0.2.0.tar.gz
  • Upload date:
  • Size: 95.8 kB
  • Tags: Source
  • Uploaded using Trusted Publishing? Yes
  • Uploaded via: twine/7.0.0 CPython/3.13.14

File hashes

Hashes for openral_reasoner-0.2.0.tar.gz
Algorithm Hash digest
SHA256 cb78d2505478fc4573f73fdb7d0d935558f6767d81c444fde7da0fdf23d2be98
MD5 2151048f29176dd2caece781266a35cc
BLAKE2b-256 0473246cdb63c4cb3e0c6cfb21bc393809347c3024f4891da39e72b3e4d0b7b7

See more details on using hashes here.

Provenance

The following attestation bundles were made for openral_reasoner-0.2.0.tar.gz:

Publisher: release-pypi.yml on OpenRAL/openral

Attestations: Values shown here reflect the state when the release was signed and may no longer be current.

File details

Details for the file openral_reasoner-0.2.0-py3-none-any.whl.

File metadata

File hashes

Hashes for openral_reasoner-0.2.0-py3-none-any.whl
Algorithm Hash digest
SHA256 b8843be87fc708da5763923f0cdc6429465a108fd13613a8b5f5ad8cf46f2ae1
MD5 ae3cc097a47af72df82b893d3d52029c
BLAKE2b-256 8f8c8eab9cd6c7c7607eba3a394532a8080f3ce1f3c19242e417049b97ed374f

See more details on using hashes here.

Provenance

The following attestation bundles were made for openral_reasoner-0.2.0-py3-none-any.whl:

Publisher: release-pypi.yml on OpenRAL/openral

Attestations: Values shown here reflect the state when the release was signed and may no longer be current.

Supported by

AWS Cloud computing and Security Sponsor Datadog Monitoring Depot Continuous Integration Fastly CDN Google Download Analytics Pingdom Monitoring Sentry Error logging StatusPage Status page