AlphaAvatar Agents
alpha-avatar-agents defines Agent execution contracts and runtime composition.
Concrete model backends and feature implementations live in plugin packages;
application hosting and transport assembly belong to Host and RTC.
📦 Package structure
alphaavatar/agents/
├── avatar/ # Engine, Loop, Context and Provider contracts.
├── runtime/ # Runtime composition, capability registry and inference.
├── configs/ # Strict Avatar and plugin-selection configuration.
├── character/ # Presentation and transport contracts.
├── memory/ # Memory interfaces and public schemas.
├── persona/ # Persona interfaces and public schemas.
├── router/ # Interaction processor contracts.
├── status/ # Status contracts.
├── tools/ # Tool abstractions and operation contracts.
├── reflection/ # Reflection contracts; not a completed feature.
├── utils/ # Workspace, identity, time and shared utilities.
├── env.py # Environment initialization.
├── log.py # Agent logging configuration.
└── version.py # Package version.
Selected files and feature directories are shown; ordinary package initializers are omitted.
Native execution
Router turn decision / direct input
-> AvatarEngine
-> Foundation Loop
-> ContextManager + ProviderService + CapabilityRegistry
-> OutputRuntime
AvatarEngine owns bounded Turn admission, duplicate/stale-input isolation, submission and precise interruption. It does not inherit a LiveKit Agent, construct model prompts, register SDK tools or write directly to Memory.
Loop owns model/tool execution and canonical record publication. ContextManager owns bounded history and query preparation. Provider owns model clients and tracing. State and static loop budgets are captured once per query, not rewritten between model steps. Generation completion remains distinct from playback completion and interruption.
Ownership
Provider schemas, enums and abstractions are in alphaavatar.agents.avatar.provider.
Plugin implementations must not introduce SDK types into Core or public output data.
Runtime components use Core LifecycleComponent and LifecycleGroup where appropriate.
AvatarRuntime composes Session, State, Core runtimes, Foundation, capabilities and its
inference client; it is not the plugin lifecycle manager.
Engine starts its consumers before Router enables decisions. It stops admission and settles Loop execution while its canonical record consumers still exist. Plugin finalization remains in each plugin. Host owns the surrounding execution lifecycle; transport detachment releases RTC subscriptions independently of Engine destruction.
InferenceRuntime owns the shared endpoint and serial runner processes. InferenceExecutor owns only client connections. Host supervises the service but does not create a second owner for runner processes. Neither failed jobs nor state-changing requests are replayed automatically by the Host.
Configuration
Agent configuration selects plugins and processor-owned init_config settings. The
main model is loop.init_config.model; there is no SDK llm section. Tool registration
does not grant approval: exact capability policy or an Engine authorizer is required.
The native provider extras are selected on alpha-avatar-plugins-provider, not on an
Agents SDK providers extra.
Host loads the shared application file and removes host before strictly validating
AvatarConfig. Execution children restore a validated snapshot, retaining the original
Avatar identity and resolved workspace rather than appending the Avatar ID again.
The alphaavatar command and its argument parser belong exclusively to Host.
Boundaries
Native TTS and Character use AlphaAvatar contracts, not AgentSession. Host supplies Character's scoped RTC transport, and the Character plugin owns browser presentation. Real renderer, RTC and provider integration still require end-to-end validation.
Session-scoped Perception/Output consistency checks remain unchanged. Persistent Episodes, cross-attachment recovery, sleep/wake and Dream scheduling are separate work.
Metadata
Release files for alpha-avatar-agents 0.6.8
For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.
Source distribution (sdist)
| File | Size | Uploaded | Provenance | |
|---|---|---|---|---|
| alpha_avatar_agents-0.6.8.tar.gz | 94.7 kB | Without provenance | Details |
Built distribution (wheel)
| File | Interpreter | ABI | Platform | Reset |
|---|---|---|---|---|
| alpha_avatar_agents-0.6.8-py3-none-any.whl | Python 3 | none | any | Details |
Total release size: 315.7 kB
Release files / alpha_avatar_agents-0.6.8.tar.gz
| Download URL | alpha_avatar_agents-0.6.8.tar.gz |
|---|---|
| Size | 94.7 kB |
| Tags | Source |
|
SHA-256 checksum How to use checksums |
1ae9094467ea3a335aa7aeceb2f7ea4f5418462b821e5872a7f57ce5cd603546
|
|
BLAKE2b-256 checksum How to use checksums |
a0cb14f7f2325853516b6956f78018bf58acd0c2cd5ef69dd83b2267f11cd267
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/7.0.0 CPython/3.13.16
|
Release files / alpha_avatar_agents-0.6.8-py3-none-any.whl
| Download URL | alpha_avatar_agents-0.6.8-py3-none-any.whl |
|---|---|
| Size | 221.0 kB |
| Tags | Python 3 |
|
SHA-256 checksum How to use checksums |
b875d69b96674ec35e62d5749841c266e3c9cae374762f8cf2b5c6ee6afe7f9e
|
|
BLAKE2b-256 checksum How to use checksums |
ed13b7774020f98f4962e09983c25b1cbe7c6c7fdfd4efe94b117534e0b964c5
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/7.0.0 CPython/3.13.16
|
Security for alpha-avatar-agents 0.6.8
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
No provenance found
None of the files in this release were uploaded with provenance metadata.
A note on package security
PyPI is a repository for distributing packages, not a service for auditing them. While third-party security researchers and our staff do their best to detect and remove malicious packages, PyPI cannot guarantee the quality, security or suitability of any given package.
To learn more about protecting yourself from malicious actors, visit our package security documentation.