Skip to main content
Apollo-1’s symbolic reasoning operates on four foundational concepts: entities, parameters, tools, and constraints. Together, they form the structured language through which agents understand and act on the world.

Entities

Entities are the structured objects in your domain — the “nouns” of your agent’s world. Examples include:
  • User Profile — customer identity and account information
  • Transactions — payment and purchase records
  • Credit Cards — financial instruments and their attributes
  • Products — items or services available
Entities are automatically populated from API responses or can be manually defined.

Parameters

Parameters are entity attributes — the “adjectives” that describe them. Each parameter includes:
  • Name — identifier for the parameter
  • Type — string, number, date, enum, or boolean
  • Description — what the parameter represents
  • Value — current or expected value
  • Use — input, output, or both

Operators by Type

Available operators vary by parameter type:
TypeOperators
StringIs, Is not, Contains, Has any value, About, Does not have any value
NumberMore than, Less than, Between, Equals, Has any value
EnumIs, Is not, Has any value, Is empty
DateBefore, After, Between, Has any value, Is, Is not
BooleanIs true, Is false

Tools

Tools are containers of logic that perform specific functions. Each tool encompasses:
  • Goal — what the tool is designed to achieve
  • Activation logic — conditions under which the tool becomes available
  • Required information — parameters needed before execution
  • Integrations — external API connections the tool uses
  • Constraints — rules governing the tool’s behavior
  • Response configuration — how results are presented to the user

Constraints

Constraints are WHEN/THEN rules that enforce business policies at the architectural level. There are three types:

Policy Constraints

Business rules that block or modify behavior based on state.
“Block disputes for transactions older than 8 days”

Confirmation Constraints

Rules requiring explicit user consent before proceeding.
“Ask for confirmation before processing any payment”

Authentication Constraints

Requirements for identity verification before sensitive operations.
“Require ID verification for refunds over $200”
Unlike instructions an LLM “tries” to follow, constraints are enforced at the architectural level. Once state is correctly interpreted, constraint enforcement is deterministic.