| name | jnana-reasoning |
| description | Scientific reasoning via Jnana CoScientist — hypothesis generation, evaluation, and parameter bounding |
| metadata | {"openclaw":{"requires":{"env":["OPENAI_API_KEY"],"bins":["python3"],"anyBins":["python3.11","python3.12"]},"primaryEnv":"OPENAI_API_KEY"},"dependencies":["jnana","academy"]} |
Jnana Reasoning — Scientific Intelligence Layer
Access Jnana's CoScientist for hypothesis-driven scientific reasoning.
This is Layer 2 of the 4-layer hybrid architecture.
Capabilities
- Research goal setting: Initialize a scientific investigation with a research goal
- Hypothesis generation: Generate testable hypotheses using literature and domain knowledge
- Next action recommendation (Tier 1): Evaluate current state, recommend next task type
- Parameter bounding (Tier 2): Generate bounded parameter configurations for computational skills
- Result evaluation: Evaluate experimental results against hypotheses via Artifact DAG
- Convergence detection: Determine when sufficient evidence has been gathered
Two-Tier Prompting
The reasoning bridge uses a two-tier prompting strategy:
-
Tier 1 — Recommendation: recommend_next_action() evaluates the current state
and recommends the next task type (computational_design, molecular_dynamics, analysis,
free_energy, or stop).
-
Tier 2 — Parameter Bounding: bound_parameters() takes the recommendation and
generates a bounded parameter configuration for the selected skill.
MCP Endpoints
jnana.set_research_goal — Initialize reasoning for a research goaljnana.recommend_next_action — Get next recommended task type (Tier 1)jnana.bound_parameters — Get parameter config for a skill (Tier 2)jnana.evaluate_results — Evaluate artifacts against hypothesesjnana.check_convergence — Check if research goal is met
Script Usage
python skills/jnana-reasoning/scripts/reason.py --help
python skills/jnana-reasoning/scripts/reason.py set-goal "Design a binder for target X"
python skills/jnana-reasoning/scripts/reason.py recommend --previous-run starting
python skills/jnana-reasoning/scripts/reason.py bound-params --skill bindcraft --task-type computational_design
python skills/jnana-reasoning/scripts/reason.py evaluate --artifact-ids abc123 def456
python skills/jnana-reasoning/scripts/reason.py check-convergence
Parameters
action: Reasoning action — "set_goal", "recommend", "bound_params", "evaluate_results", "check_convergence"research_goal: Research goal description (for "set_goal")previous_run_type: Previous run type for context (for "recommend")skill_name: Target skill for parameter recommendations (for "bound_params")task_type: Task type for parameter schema (for "bound_params")artifact_ids: Artifact IDs to evaluate (for "evaluate_results")
