itt-field-store 0.2.0

Intent Tensor Theory — Field-based compute substrate replacing SQL

ITT Field Store

Intent Tensor Theory — Field-based compute substrate replacing SQL

"Topological sort was the right solution for the hardware of 1979. The dependency chain is an unnecessary constraint." — WP-06

---

The idea

SQL runs queries sequentially on relational tables.
ITT runs queries simultaneously on a living field.

Instead of SELECT * FROM users WHERE role='admin', you inject an intent into the field and read which nodes activate above a threshold. Circular dependencies aren't errors — they're fixed points resolved by Banach contraction.

Math: Graph Laplacian Diffusion + Allen-Cahn Phase Separation + Banach Fixed-Point Convergence
Reference: WP-06: Death of the Dependency Chain

---

Install

pip install itt-field-store

With API server:

pip install itt-field-store[api]

---

Usage

Local (embedded, like SQLite)

from itt import FieldStore

store = FieldStore("my_store")

# Insert (replaces INSERT INTO)
store.table("users").insert([
    {"_id": "1", "name": "Alice", "role": "admin", "active": True},
    {"_id": "2", "name": "Bob",   "role": "user",  "active": True},
    {"_id": "3", "name": "Carol", "role": "admin", "active": False},
])

# Query (replaces SELECT * WHERE)
results = store.table("users").intent({"role": "admin"}).top(10).fetch()
for r in results:
    print(r["name"], r["_phi"])   # _phi is the field activation score

Stateful living field (the real ITT mode)

from itt import DeltaState

state = DeltaState("production_field")

# Absorb new data — field evolves, doesn't reset
state.absorb(new_records)

# Query with semantic intent
result = state.query("find all active administrators")

# Results above threshold
print(result.above_threshold(0.4))

# Convergence metadata
print(result.convergence_report())

# Anomaly detection — nodes in semantic tension
print(result.instability_mask())

# Persist
state.save("./my_field.itt")
state = DeltaState.load("./my_field.itt")

Remote client (like Supabase)

from itt import ITTClient

client = ITTClient("https://intent-tensor-theory-api.hf.space")

client.table("users").insert([{"name": "Alice", "role": "admin"}])
results = client.table("users").query({"role": "admin"}).top(5).fetch()

MCP Tool (callable by Claude, GPT, any LLM)

# Register in your LLM client
tools = client.tools()   # returns MCP tool definitions

# Or run the MCP server:
# python -m itt.mcp.server

---

SQL → ITT mapping

SQL	ITT
CREATE TABLE	store.table("name") (no schema needed)
INSERT INTO	.insert(records)
SELECT * WHERE	.intent({...}).fetch()
SELECT * LIMIT n	.top(n).fetch()
UPDATE SET WHERE	.upsert(id, patch)
DELETE WHERE	.delete([ids])
Circular reference → ERROR	Fixed point → converges
Sequential evaluation	Simultaneous field update
No anomaly detection	.instability_mask()

---

Deploy to HuggingFace Spaces

# Clone the repo, push to a new HF Space
git clone https://github.com/intent-tensor-theory/itt-field-store
cd itt-field-store
# push to HF Space → public API at your-space.hf.space

---

intent-tensor-theory.com · Coordinate System · Code Equations