ouroboros-ai 0.25.3.dev3

Self-Improving AI Workflow System

English | 한국어

◯ ─────────── ◯



O U R O B O R O S

◯ ─────────── ◯

Stop prompting. Start specifying.
_{A Claude Code plugin that turns vague ideas into validated specs — before AI writes a single line of code.}

Quick Start · Philosophy · How · Commands · Agents

---

Security Notice (2026-03-24): LiteLLM PyPI Supply Chain Attack

LiteLLM versions 1.82.7 and 1.82.8 on PyPI contained a malicious .pth file that auto-executes on every Python startup (no import needed), exfiltrating environment variables, API keys, SSH keys, and cloud credentials to an external server.

Am I affected? You are at risk if, after March 23, 2026, you either:

• Fresh-installed Ouroboros (pip install ouroboros-ai, uv tool install ouroboros-ai), or
• Upgraded an existing install (pip install --upgrade ouroboros-ai)

In both cases, litellm>=1.80.0 was a dependency, and your package manager would have resolved it to the malicious 1.82.7/1.82.8. If you installed before that date and haven't reinstalled or upgraded since, you are likely safe.

Ouroboros has removed litellm entirely as of v0.25.2 (stable) and v0.26.0b7 (beta). However, upgrading Ouroboros only removes litellm from the dependency list — it does NOT uninstall litellm itself or delete the malicious .pth file already on your system. You must clean up manually:

# 1. Find the malicious file — delete any results
find / -name "litellm_init.pth" 2>/dev/null

# 2. Clean package manager caches
uv cache clean litellm    # if using uv
pip cache purge            # if using pip

# 3. Rotate ALL credentials that were on your system during the exposure window

References: LiteLLM #24512 · Ouroboros #195 · Analysis

---

AI can build anything. The hard part is knowing what to build.

Ouroboros is a specification-first AI development system. It applies Socratic questioning and ontological analysis to expose your hidden assumptions — before a single line of code is written.

Most AI coding fails at the input, not the output. The bottleneck isn't AI capability. It's human clarity. Ouroboros fixes the human, not the machine.

---

From Wonder to Ontology

Wonder → "How should I live?" → "What IS 'live'?" → Ontology — Socrates

This is the philosophical engine behind Ouroboros. Every great question leads to a deeper question — and that deeper question is always ontological: not "how do I do this?" but "what IS this, really?"

   Wonder                          Ontology
     💡                               🔬
"What do I want?"    →    "What IS the thing I want?"
"Build a task CLI"   →    "What IS a task? What IS priority?"
"Fix the auth bug"   →    "Is this the root cause, or a symptom?"

This is not abstraction for its own sake. When you answer "What IS a task?" — deletable or archivable? solo or team? — you eliminate an entire class of rework. The ontological question is the most practical question.

Ouroboros embeds this into its architecture through the Double Diamond:

    ◇ Wonder          ◇ Design
   ╱  (diverge)      ╱  (diverge)
  ╱    explore      ╱    create
 ╱                 ╱
◆ ──────────── ◆ ──────────── ◆
 ╲                 ╲
  ╲    define       ╲    deliver
   ╲  (converge)     ╲  (converge)
    ◇ Ontology        ◇ Evaluation

The first diamond is Socratic: diverge into questions, converge into ontological clarity. The second diamond is pragmatic: diverge into design options, converge into verified delivery. Each diamond requires the one before it — you cannot design what you haven't understood.

---

Quick Start

Step 1 — Install the plugin (in your terminal):

claude plugin marketplace add Q00/ouroboros
claude plugin install ouroboros@ouroboros

Step 2 — Run setup (inside a Claude Code session):

# Start Claude Code, then type:
ooo setup

ooo commands are Claude Code skills — they run inside a Claude Code session, not in your terminal. Setup registers the MCP server globally (one-time) and optionally adds an Ouroboros reference block to your project's CLAUDE.md.

Step 3 — Start building:

ooo interview "I want to build a task management CLI"

What just happened?

ooo interview  →  Socratic questioning exposed 12 hidden assumptions
ooo seed       →  Crystallized answers into an immutable spec (Ambiguity: 0.15)
ooo run        →  Executed via Double Diamond decomposition
ooo evaluate   →  3-stage verification: Mechanical → Semantic → Consensus

The serpent completed one loop. Each loop, it knows more than the last.

---

The Loop

The ouroboros — a serpent devouring its own tail — isn't decoration. It IS the architecture:

    Interview → Seed → Execute → Evaluate
        ↑                           ↓
        └──── Evolutionary Loop ────┘

Each cycle doesn't repeat — it evolves. The output of evaluation feeds back as input for the next generation, until the system truly knows what it's building.

Phase	What Happens
Interview	Socratic questioning exposes hidden assumptions
Seed	Answers crystallize into an immutable specification
Execute	Double Diamond: Discover → Define → Design → Deliver
Evaluate	3-stage gate: Mechanical ($0) → Semantic → Multi-Model Consensus
Evolve	Wonder ("What do we still not know?") → Reflect → next generation

"This is where the Ouroboros eats its tail: the output of evaluation becomes the input for the next generation's seed specification." — reflect.py

Convergence is reached when ontology similarity ≥ 0.95 — when the system has questioned itself into clarity.

Ralph: The Loop That Never Stops

ooo ralph runs the evolutionary loop persistently — across session boundaries — until convergence is reached. Each step is stateless: the EventStore reconstructs the full lineage, so even if your machine restarts, the serpent picks up where it left off.

Ralph Cycle 1: evolve_step(lineage, seed) → Gen 1 → action=CONTINUE
Ralph Cycle 2: evolve_step(lineage)       → Gen 2 → action=CONTINUE
Ralph Cycle 3: evolve_step(lineage)       → Gen 3 → action=CONVERGED ✓
                                                └── Ralph stops.
                                                    The ontology has stabilized.

Ambiguity Score: The Gate Between Wonder and Code

The Interview doesn't end when you feel ready — it ends when the math says you're ready. Ouroboros quantifies ambiguity as the inverse of weighted clarity:

Ambiguity = 1 − Σ(clarityᵢ × weightᵢ)

Each dimension is scored 0.0–1.0 by the LLM (temperature 0.1 for reproducibility), then weighted:

Dimension	Greenfield	Brownfield
Goal Clarity — Is the goal specific?	40%	35%
Constraint Clarity — Are limitations defined?	30%	25%
Success Criteria — Are outcomes measurable?	30%	25%
Context Clarity — Is the existing codebase understood?	—	15%

Threshold: Ambiguity ≤ 0.2 — only then can a Seed be generated.

Example (Greenfield):

  Goal: 0.9 × 0.4  = 0.36
  Constraint: 0.8 × 0.3  = 0.24
  Success: 0.7 × 0.3  = 0.21
                        ──────
  Clarity             = 0.81
  Ambiguity = 1 − 0.81 = 0.19  ≤ 0.2 → ✓ Ready for Seed

Why 0.2? Because at 80% weighted clarity, the remaining unknowns are small enough that code-level decisions can resolve them. Above that threshold, you're still guessing at architecture.

Ontology Convergence: When the Serpent Stops

The evolutionary loop doesn't run forever. It stops when consecutive generations produce ontologically identical schemas. Similarity is measured as a weighted comparison of schema fields:

Similarity = 0.5 × name_overlap + 0.3 × type_match + 0.2 × exact_match

Component	Weight	What It Measures
Name overlap	50%	Do the same field names exist in both generations?
Type match	30%	Do shared fields have the same types?
Exact match	20%	Are name, type, AND description all identical?

Threshold: Similarity ≥ 0.95 — the loop converges and stops evolving.

But raw similarity isn't the only signal. The system also detects pathological patterns:

Signal	Condition	What It Means
Stagnation	Similarity ≥ 0.95 for 3 consecutive generations	Ontology has stabilized
Oscillation	Gen N ≈ Gen N-2 (period-2 cycle)	Stuck bouncing between two designs
Repetitive feedback	≥ 70% question overlap across 3 generations	Wonder is asking the same things
Hard cap	30 generations reached	Safety valve

Gen 1: {Task, Priority, Status}
Gen 2: {Task, Priority, Status, DueDate}     → similarity 0.78 → CONTINUE
Gen 3: {Task, Priority, Status, DueDate}     → similarity 1.00 → CONVERGED ✓

Two mathematical gates, one philosophy: don't build until you're clear (Ambiguity ≤ 0.2), don't stop evolving until you're stable (Similarity ≥ 0.95).

---

Commands

All ooo commands run inside a Claude Code session, not in your terminal. Run ooo setup after installation to register the MCP server (one-time) and optionally integrate with your project's CLAUDE.md.

Command	What It Does
ooo setup	Register MCP server (one-time)
ooo interview	Socratic questioning → expose hidden assumptions
ooo seed	Crystallize into immutable spec
ooo run	Execute via Double Diamond decomposition
ooo evaluate	3-stage verification gate
ooo evolve	Evolutionary loop until ontology converges
ooo unstuck	5 lateral thinking personas when you're stuck
ooo status	Drift detection + session tracking
ooo ralph	Persistent loop until verified
ooo tutorial	Interactive hands-on learning
ooo help	Full reference

---

The Nine Minds

Nine agents, each a different mode of thinking. Loaded on-demand, never preloaded:

Agent	Role	Core Question
Socratic Interviewer	Questions-only. Never builds.	"What are you assuming?"
Ontologist	Finds essence, not symptoms	"What IS this, really?"
Seed Architect	Crystallizes specs from dialogue	"Is this complete and unambiguous?"
Evaluator	3-stage verification	"Did we build the right thing?"
Contrarian	Challenges every assumption	"What if the opposite were true?"
Hacker	Finds unconventional paths	"What constraints are actually real?"
Simplifier	Removes complexity	"What's the simplest thing that could work?"
Researcher	Stops coding, starts investigating	"What evidence do we actually have?"
Architect	Identifies structural causes	"If we started over, would we build it this way?"

---

Under the Hood

18 packages · 166 modules · 95 test files · Python 3.14+

src/ouroboros/
├── bigbang/        Interview, ambiguity scoring, brownfield explorer
├── routing/        PAL Router — 3-tier cost optimization (1x / 10x / 30x)
├── execution/      Double Diamond, hierarchical AC decomposition
├── evaluation/     Mechanical → Semantic → Multi-Model Consensus
├── evolution/      Wonder / Reflect cycle, convergence detection
├── resilience/     4-pattern stagnation detection, 5 lateral personas
├── observability/  3-component drift measurement, auto-retrospective
├── persistence/    Event sourcing (SQLAlchemy + aiosqlite), checkpoints
├── orchestrator/   Claude Agent SDK integration, session management
├── core/           Types, errors, seed, ontology, security
├── providers/      LiteLLM adapter (100+ models)
├── mcp/            MCP client/server for Claude Code
├── plugin/         Claude Code plugin system
├── tui/            Terminal UI dashboard
└── cli/            Typer-based CLI

Key internals:

• PAL Router — Frugal (1x) → Standard (10x) → Frontier (30x) with auto-escalation on failure, auto-downgrade on success
• Drift — Goal (50%) + Constraint (30%) + Ontology (20%) weighted measurement, threshold ≤ 0.3
• Brownfield — Scans 15 config file types across 12+ language ecosystems
• Evolution — Up to 30 generations, convergence at ontology similarity ≥ 0.95
• Stagnation — Detects spinning, oscillation, no-drift, and diminishing returns patterns

---

Real-Time Monitoring (TUI)

Ouroboros includes a terminal dashboard for real-time workflow monitoring. Run it in a separate terminal while ooo run or ooo evolve is executing:

# Install and launch
uvx --from ouroboros-ai ouroboros tui monitor

# Or if installed locally
uv run ouroboros tui monitor

Key	Screen	What You See
1	Dashboard	Phase progress, acceptance criteria tree, live status
2	Execution	Timeline, phase outputs, detailed events
3	Logs	Filterable log viewer with level-based coloring
4	Debug	State inspector, raw events, configuration

See TUI Usage Guide for full details.

---

Contributing

git clone https://github.com/Q00/ouroboros
cd ouroboros
uv sync --all-groups && uv run pytest

Issues · Discussions

---

Star History

---

"The beginning is the end, and the end is the beginning."

The serpent doesn't repeat — it evolves.

MIT License