Domain Types¶

MOL has 8 built-in domain types designed for cognitive computing and AI applications.

Base: MolObject¶

Every domain type inherits from MolObject, which provides:

_id — unique 8-character hex identifier
_created_at — creation timestamp
_access_level — security level (default: "public")

Thought¶

An idea with a confidence score.

let idea be Thought("Neural networks learn patterns", 0.87)
show idea.content       -- "Neural networks learn patterns"
show idea.confidence    -- 0.87
show type_of(idea)      -- "Thought"

Fields:

Field	Type	Description
`content`	Text	The thought content
`confidence`	Number	0.0 — 1.0 confidence score
`tags`	List	Tags (added via `.tag()`)
`linked_thoughts`	List	Linked thoughts

Operations:

-- Use in pipes with think()
let response be "AI is transforming healthcare"
  |> think("analyze this statement")
show response.confidence

Memory¶

Labeled storage with strength decay.

let mem be Memory("session_key", "important data")
show mem.key            -- "session_key"
show mem.value          -- "important data"
show mem.strength       -- 1.0 (initial)

Fields:

Field	Type	Description
`key`	Text	Memory label
`value`	any	Stored value
`strength`	Number	Recall strength (decays over time)
`recall_count`	Number	Times recalled

Operations:

let value be recall(mem)    -- Boosts strength by 0.1

Node¶

A neural network node with connections.

let neuron be Node("cortex", 0.75)
show neuron.label       -- "cortex"
show neuron.weight      -- 0.75
show neuron.generation  -- 0

Fields:

Field	Type	Description
`label`	Text	Node name
`weight`	Number	Connection weight
`connections`	List	Connected nodes
`active`	Bool	Active state
`generation`	Number	Evolution generation

Operations:

let a be Node("input", 0.5)
let b be Node("hidden", 0.8)
link a to b              -- Connect nodes
evolve a                 -- weight *= 1.1, generation += 1

Stream¶

Event stream with publish/subscribe.

let s be Stream("data_feed")
listen "data_ready" do
  show "Data arrived!"
end
emit "data_ready"        -- Triggers listener

Document¶

A source document for pipeline processing.

let doc be Document("paper.txt", "Machine learning is...")
show doc.source          -- "paper.txt"
show doc.content         -- "Machine learning is..."

Fields:

Field	Type	Description
`source`	Text	File name or URL
`content`	Text	Document text
`metadata`	Map	Additional metadata

Use in pipes:

let index be Document("data.txt", text)
  |> chunk(512)
  |> embed("model-v1")
  |> store("index")

Chunk¶

A segment of a document.

let chunks be chunk(doc, 100)
show len(chunks)
show chunks[0].content
show chunks[0].index     -- 0
show chunks[0].source    -- "paper.txt"

Embedding¶

A vector representation of text.

let emb be Embedding("hello world", "mol-sim-v1")
show emb.text            -- "hello world"
show emb.model           -- "mol-sim-v1"
show emb.dimensions      -- 64

Deterministic Vectors

MOL generates deterministic 64-dimensional pseudo-embeddings using SHA-256 hashing. Same text always produces the same vector — useful for testing and reproducible pipelines.

VectorStore¶

A searchable index of embeddings with cosine similarity search.

-- Build an index
let index be doc |> chunk(100) |> embed("v1") |> store("kb")

-- Search
let results be retrieve("query text", "kb", 3)
for r in results do
  show r.text + " (score: " + to_text(r.score) + ")"
end

Fields:

Field	Type	Description
`name`	Text	Store name
`entries`	List	Stored vectors

Type Checking¶

show type_of(Thought("x", 0.5))        -- "Thought"
show type_of(Document("f", "content"))  -- "Document"
show type_of(Node("n", 0.5))           -- "Node"

-- Type-annotated declarations
let t: Thought be Thought("idea", 0.9)  -- ✓