> ## Documentation Index > Fetch the complete documentation index at: https://praison.ai/docs/llms.txt > Use this file to discover all available pages before exploring further. # Standalone LLM Modules > Model catalogue, endpoint resolver, credentials gate, and AutoGen config-list — usable with pip install praisonai-code alone After the C7 milestone, the LLM helpers live in `praisonai-code` — no wrapper import required. ```python theme={"theme":{"light":"vitesse-light","dark":"vitesse-dark"}} from praisonaiagents import Agent from praisonai_code.llm.catalogue import ModelCatalogue from praisonai_code.llm.credentials import is_configured catalogue = ModelCatalogue() models = catalogue.list_models(provider="openai") if is_configured(): agent = Agent( name="LLM Info Agent", instructions=f"You have access to {len(models)} OpenAI models.", ) agent.start("Which OpenAI model supports vision and tool calling?") ``` The user asks about models; catalogue and credential helpers from `praisonai-code` answer without wrapper imports. ```mermaid theme={"theme":{"light":"vitesse-light","dark":"vitesse-dark"}} graph LR subgraph "Standalone LLM Modules" Ask[📋 Ask about models] --> Cat[📚 Catalogue + Credentials] Cat --> Agent[🤖 Agent] Agent --> Ans[✅ Answer] end classDef input fill:#6366F1,stroke:#7C90A0,color:#fff classDef module fill:#189AB4,stroke:#7C90A0,color:#fff classDef agent fill:#8B0000,stroke:#7C90A0,color:#fff classDef output fill:#10B981,stroke:#7C90A0,color:#fff class Ask input class Cat module class Agent agent class Ans output ``` ## How It Works ```mermaid theme={"theme":{"light":"vitesse-light","dark":"vitesse-dark"}} sequenceDiagram participant User participant Agent participant Credentials as llm.credentials participant Catalogue as llm.catalogue User->>Agent: Ask about available models Agent->>Credentials: is_configured() Credentials-->>Agent: Ready Agent->>Catalogue: list_models(provider="openai") Catalogue-->>Agent: Model list Agent-->>User: Answer with model details ``` ### Module Map ```mermaid theme={"theme":{"light":"vitesse-light","dark":"vitesse-dark"}} graph TB subgraph "pip install praisonai-code" CAT[📚 llm.catalogue] CFG[⚙️ llm.config] ENV[🔍 llm.env] CRD[🔑 llm.credentials] end subgraph "Exports" CAT --> MC[ModelCatalogue\nModelInfo] CFG --> BCL[build_config_list] ENV --> EP[LLMEndpoint\nresolve_llm_endpoint\ndefault_model_for_available_provider] CRD --> IC[is_configured\ninject_credentials_into_env\nresolve_llm_endpoint_with_credentials] end subgraph "praisonai.llm.* shim" SH[PEP 562 sys.modules alias] SH -.->|same object| CAT SH -.->|same object| CFG SH -.->|same object| ENV SH -.->|same object| CRD end classDef module fill:#189AB4,stroke:#7C90A0,color:#fff classDef export fill:#10B981,stroke:#7C90A0,color:#fff classDef shim fill:#F59E0B,stroke:#7C90A0,color:#fff class CAT,CFG,ENV,CRD module class MC,BCL,EP,IC export class SH shim ``` ## Quick Start ```python theme={"theme":{"light":"vitesse-light","dark":"vitesse-dark"}} from praisonaiagents import Agent from praisonai_code.llm.catalogue import ModelCatalogue catalogue = ModelCatalogue() models = catalogue.list_models(provider="openai") agent = Agent( name="LLM Info Agent", instructions=f"You have access to {len(models)} OpenAI models.", ) agent.start("Which OpenAI model supports vision and tool calling?") ``` ```python theme={"theme":{"light":"vitesse-light","dark":"vitesse-dark"}} from praisonaiagents import Agent from praisonai_code.llm.credentials import is_configured, inject_credentials_into_env if not is_configured(): raise SystemExit("Run: praisonai setup (or set OPENAI_API_KEY)") inject_credentials_into_env() agent = Agent( name="Secure Agent", instructions="You are a helpful assistant.", ) agent.start("What models are available for my account?") ``` ```python theme={"theme":{"light":"vitesse-light","dark":"vitesse-dark"}} from praisonai_code.llm.catalogue import ModelCatalogue catalogue = ModelCatalogue() models = catalogue.list_models(provider="openai") print(f"{len(models)} OpenAI models available") ``` ```python theme={"theme":{"light":"vitesse-light","dark":"vitesse-dark"}} from praisonai_code.llm.credentials import is_configured, inject_credentials_into_env from praisonai_code.llm.env import resolve_llm_endpoint if is_configured(): inject_credentials_into_env() ep = resolve_llm_endpoint() print(f"Using: {ep.model} at {ep.base_url}") ``` *** ## How It Works The user asks about models; the agent queries the catalogue and credential helpers, then answers. ```mermaid theme={"theme":{"light":"vitesse-light","dark":"vitesse-dark"}} sequenceDiagram participant User participant Agent participant Catalogue as llm.catalogue participant Credentials as llm.credentials User->>Agent: "Which model supports vision?" Agent->>Credentials: is_configured() Credentials-->>Agent: True Agent->>Catalogue: list_models(provider="openai") Catalogue-->>Agent: ModelInfo list Agent-->>User: Answer with matching model ``` *** ## The Four Modules | Module | Public exports | Purpose | | -------------------------------- | --------------------------------------------------------------------------------------- | ------------------------------------------------------ | | `praisonai_code.llm.catalogue` | `ModelCatalogue`, `ModelInfo` | Discover and validate LLM models | | `praisonai_code.llm.config` | `build_config_list` | AutoGen-style `[{model, base_url, api_key, api_type}]` | | `praisonai_code.llm.env` | `LLMEndpoint`, `resolve_llm_endpoint`, `default_model_for_available_provider` | Resolve which model/endpoint to hit | | `praisonai_code.llm.credentials` | `is_configured`, `inject_credentials_into_env`, `resolve_llm_endpoint_with_credentials` | Credential gate for CLI startup | *** ## How It Works The user asks about models; the catalogue and credential gate resolve an endpoint, then the agent answers. ```mermaid theme={"theme":{"light":"vitesse-light","dark":"vitesse-dark"}} sequenceDiagram participant User participant Agent participant Catalogue participant Credentials User->>Agent: "Which model supports vision?" Agent->>Catalogue: list_models(provider="openai") Catalogue-->>Agent: Available models Agent->>Credentials: is_configured() Credentials-->>Agent: Endpoint + key Agent-->>User: Answer with model details ``` *** ## Which Import Path Should I Use? ```mermaid theme={"theme":{"light":"vitesse-light","dark":"vitesse-dark"}} graph TB Q{"Writing new code
targeting praisonai-code?"} -->|Yes| NEW[from praisonai_code.llm.* import ...] Q -->|No — existing code with praisonai.llm.*| KEEP[Keep as-is — shim guarantees identity] LEGACY{"Have praisonai wrapper
installed?"} -->|Yes| KEEP LEGACY -->|No — standalone only| NEW classDef decision fill:#6366F1,stroke:#7C90A0,color:#fff classDef outcome fill:#10B981,stroke:#7C90A0,color:#fff class Q,LEGACY decision class NEW,KEEP outcome ``` *** ## Backward Compatibility Guarantee All four `praisonai.llm.*` module paths still resolve — a `sys.modules` alias in the wrapper points at `praisonai_code.llm.*`. The unit test `test_c5_backward_compat.py::test_module_identity` asserts: * `praisonai.llm.catalogue is praisonai_code.llm.catalogue` * `praisonai.llm.config is praisonai_code.llm.config` * `praisonai.llm.env is praisonai_code.llm.env` * `praisonai.llm.credentials is praisonai_code.llm.credentials` You do not need to migrate existing code — both paths point at the same object. *** ## Standalone Smoke Test This example runs in a venv with only `praisonai-agents` + `praisonai-code` installed (no wrapper): ```python theme={"theme":{"light":"vitesse-light","dark":"vitesse-dark"}} from praisonai_code.cli.configuration.resolver import ConfigResolver from praisonai_code._version import get_package_version from praisonai_code.llm.credentials import is_configured from praisonai_code.llm.catalogue import ModelCatalogue assert ModelCatalogue().list_models(provider="openai") # 132 models ``` *** ## Module Reference ### `praisonai_code.llm.catalogue` ```python theme={"theme":{"light":"vitesse-light","dark":"vitesse-dark"}} from praisonai_code.llm.catalogue import ModelCatalogue, ModelInfo catalogue = ModelCatalogue( cache_dir=None, # defaults to ~/.praison/cache cache_ttl=3600, # 1 hour ) # List all models, optionally filter models = catalogue.list_models(provider="openai", search="gpt-4") # Get full metadata for one model info = catalogue.describe_model("gpt-4o") # Validate a model ID — raises ValueError with suggestions on mismatch try: valid_id = catalogue.validate_model("gpt-4o-mni") except ValueError as e: print(e) # "Unknown model 'gpt-4o-mni'. Did you mean: gpt-4o-mini, gpt-4o" # Quick boolean check ok = catalogue.is_valid_model("gpt-4o") # True # Fuzzy suggestions suggestions = catalogue.get_suggestions("gpt4o") # ['gpt-4o', 'gpt-4o-mini'] ``` Cache file: `~/.praison/cache/models.json` (TTL 3600s). Falls back to 14 curated models (OpenAI 5, Anthropic 3, Google 3, Groq 2, Ollama 1) when litellm is unavailable. ### `praisonai_code.llm.config` ```python theme={"theme":{"light":"vitesse-light","dark":"vitesse-dark"}} from praisonai_code.llm.config import build_config_list config = build_config_list(include_api_type=True) # → [{'model': 'gpt-4o-mini', 'base_url': 'https://api.openai.com/v1', # 'api_key': 'sk-...', 'api_type': 'openai'}] ``` ### `praisonai_code.llm.env` ```python theme={"theme":{"light":"vitesse-light","dark":"vitesse-dark"}} from praisonai_code.llm.env import ( LLMEndpoint, resolve_llm_endpoint, default_model_for_available_provider, ) ep = resolve_llm_endpoint() print(ep.model, ep.base_url, ep.api_key) model = default_model_for_available_provider() ``` ### `praisonai_code.llm.credentials` ```python theme={"theme":{"light":"vitesse-light","dark":"vitesse-dark"}} from praisonai_code.llm.credentials import ( is_configured, inject_credentials_into_env, resolve_llm_endpoint_with_credentials, ) if is_configured(): inject_credentials_into_env() ep = resolve_llm_endpoint_with_credentials() ``` *** ## Best Practices New scripts that target the `praisonai-code` package should import from `praisonai_code.llm.*` to avoid taking a dependency on the wrapper: ```python theme={"theme":{"light":"vitesse-light","dark":"vitesse-dark"}} from praisonai_code.llm.catalogue import ModelCatalogue from praisonai_code.llm.env import resolve_llm_endpoint ``` Existing code using `from praisonai.llm.catalogue import ModelCatalogue` works without change — the shim guarantees module identity. Migration has no functional benefit. Use `is_configured()` as a fast gate before running an agent — it checks env vars and the credential store without making any network calls: ```python theme={"theme":{"light":"vitesse-light","dark":"vitesse-dark"}} from praisonai_code.llm.credentials import is_configured, inject_credentials_into_env if not is_configured(): print("Run: praisonai setup (or set OPENAI_API_KEY)") raise SystemExit(1) inject_credentials_into_env() ``` *** ## Configuration Options Python configuration options for ModelCatalogue Python configuration options for LLMEndpoint *** ## Related CLI commands for models list, describe, validate Build AutoGen config\_list from the resolved endpoint Endpoint precedence and provider routing The standalone CLI runtime these modules power Three-tier ownership and the `_wrapper_bridge` pattern