Crate agcodex_core

Expand description

Root of the codex-core library.

Modules§

code_tools: Unified code tools scaffolding for AGCodex.
codex
config
config_profile
config_types: Types used to define the fields of crate::config::Config.
context_engine: AGCodex Context Engine scaffolding.
conversation: Conversation management utilities
embeddings: Independent embeddings system - completely separate from chat/LLM models.
embeddings_capability: Capability checks for embeddings providers. Per policy: OpenAI embeddings require an OpenAI API key. ChatGPT login (web tokens) is not sufficient and does not grant embeddings access.
error
exec
exec_env
git_info
landlock
model_family
models
modes: Operating modes for AGCodex (Plan, Build, Review). Minimal scaffolding to start the refactor without impacting existing flows.
parse_command
plan_tool
protocol: Defines the protocol for a Codex session between a client and an agent.
protocol_config_types
seatbelt
shell
spawn
subagents: Subagent system for AGCodex
tools: Internal agent tools for AGCodex
turn_diff_tracker
user_agent
util

CodexConversation
ConversationManager: ConversationManager is responsible for creating conversations and maintaining them in memory.
ModelProviderInfo: Serializable representation of a provider definition.
NewConversation: Represents a newly created Codex conversation, including the first event (which is EventMsg::SessionConfigured).

WireApi: Wire protocol that the provider speaks. Most third-party services only implement the classic OpenAI Chat Completions JSON schema, whereas OpenAI itself (and a handful of others) additionally expose the more modern Responses API. The two protocols use different request/response shapes and cannot be auto-detected at runtime, therefore each provider entry must declare which one it expects.