Skip to content

Episteme Studio Workbench

Episteme Studio is an interactive visual workbench, orchestration pipeline, and analytical environment for computational epistemology, structuralist philosophy of science, and argument mining.

Quick Launch & Operating Modes

Episteme Studio can be launched immediately via the monorepo CLI.

Instant Demo Mode (Zero Setup Required)

To explore theory graphs and interface features without connecting to Neo4j or an LLM gateway, launch the canned demonstration server:

uv run episteme-studio serve --demo

Once running, navigate to http://127.0.0.1:8000 in your browser. Demo mode populates the explorer with pre-computed scientific runs, epistemic metrics, and graph diffs.

Production Mode (Live Database Binding)

To inspect your actual pipeline runs and query your live Neo4j 5.x database:

uv run episteme-studio serve --host 127.0.0.1 --port 8000

CLI Command Options

options:
  -h, --help            show this help message and exit
  --host HOST           Host interface to bind (default: 127.0.0.1)
  --port PORT           Port to listen on (default: 8000)
  --demo                Launch with canned demonstration runs and mock dependencies
  --runs-dir DIR        Path to directory containing run manifests (default: .pipeline_runs)
  --artifacts-dir DIR   Path to directory containing artifact envelopes (default: .pipeline_artifacts)
  --dev [URL]           Proxy frontend requests to Vite dev server (default: http://127.0.0.1:5173)
  --token TOKEN         Authentication token (required when binding to external hosts)

Core Workbench Views

Episteme Studio organizes epistemic analysis into four primary views:

flowchart TD
    Workbench[Episteme Studio Workbench] --> Explorer[1. Explorer View]
    Workbench --> Execution[2. Execution View]
    Workbench --> CypherConsole[3. Cypher Console]
    Workbench --> Comparison[4. Run Comparison Engine]

    Explorer -->|Visualize| Layers[L1 Sources, L2 Concepts, L3 TheoryNet]
    Execution -->|Compose & Launch| DAG[Pre-Flight Checks & Cost Forecasting]
    CypherConsole -->|Query| Sandbox[Read-Only Ad-Hoc Cypher Playground]
    Comparison -->|Diff Runs| Deltas[Set-Theoretic Deltas & KDE Shifts]

View 1: Multi-Layer Theory Canvas (Explorer)

The main canvas renders the generated theory graph using a GPU-accelerated force-directed layout:

Graph Element Visual Encoding Epistemic Function
Layer 1 (L1 Sources) Blue nodes Ground-truth academic sources and text chunk anchors.
Layer 2 (L2 Concepts) Green nodes Extracted domain concepts, named entities, and empirical terms.
Layer 3 (L3 TheoryNet) Purple nodes Higher-level theoretical constructs, latent hypotheses, and structural laws.
Dialectical Support Green directed vectors Formal argumentative warrant: SUPPORT relations.
Dialectical Attack Red directed vectors Formal argumentative objections: ATTACK relations.
Unscoped / Drift Orange nodes Entities or assertions operating outside axiomatic boundaries.

Canvas Controls

  • Global Entity Search (⌘K): Instant fuzzy lookup across all extracted concepts and theoretical terms.
  • Epistemic Lenses: Filter the visible canvas according to structuralist partitions:
  • Empirical Base (\(M_{pp}\)): Isolates non-theoretical observations, empirical data, and grounded claims.
  • Theoretical Postulates (\(M\)): Isolates abstract axioms, latent variables, and structural laws (Stegmüller 1976 / Balzer 1987).

View 2: Pipeline Orchestrator (Execution)

The Execution view provides interactive DAG composition: * Thinking Level Switch: Regulate model reasoning compute (Off, Low, Medium, High, Custom) for complex relation extraction. * Pre-Flight Readiness Inspection: Static configuration validator verifying Neo4j index status and endpoint responsiveness prior to dispatch. * Cost & Latency Forecasting: Real-time pre-execution estimates calculating anticipated compute duration and API invocation costs. * Reproducible CLI Generator: Generates headless bash commands corresponding to active UI configurations for remote compute cluster execution.

View 3: Safe Database Playground (Cypher Console)

  • Safety Sandbox: Enforces a strict Read-Only mode to protect generated theoretical structures, embeddings, and relations from accidental deletion during exploratory queries.
  • Sample Queries Catalog: Quick-insert templates for centrality, bridging concepts, and argument chain inspection.
  • Keyboard Shortcut Dispatch (⌘↩): Execute ad-hoc structural queries directly against the active graph namespace.

View 4: Run Comparison & Set-Theoretic Deltas (Comparison)

Evaluate how varying prompt strategies, temperature settings, or underlying model architectures alter theory reconstruction:

  • Visual Canvas Diff: Color-coded node and edge highlighting showing structural growth, retention, and pruning between Run A and Run B.
  • Kernel Density Estimation (KDE): Overlapping density curves visualizing shifts in semantic extraction confidence across runs.
  • Set-Theoretic Delta Cardinality:
  • Intersection (\(A \cap B\)): Invariant entities and edges verified across both configurations.
  • Pruned Set (\(A \setminus B\)): Baseline structural elements omitted by the candidate configuration.
  • Discovered Set (\(B \setminus A\)): Novel entities and dialectical relations introduced by the candidate run.

Next Steps