Skip to content

🌟 The Multi-Agent Framework: First AI Software Company, Towards Natural Language Programming

Add Llama API Support

cmodi-meta43b090c472ed
diff --git a/config/examples/llamaapi-llama-4-Maverick-17B-128E-Instruct-FP8.yaml b/config/examples/llamaapi-llama-4-Maverick-17B-128E-Instruct-FP8.yaml
new file mode 100644
index 000000000..9b48d770c
--- /dev/null
+++ b/config/examples/llamaapi-llama-4-Maverick-17B-128E-Instruct-FP8.yaml
@@ -0,0 +1,5 @@
+llm:
+ api_type: llamaapi
+ base_url: "https://api.llama.com/compat/v1/"
+ api_key: "YOUR_API_KEY"
+ model: llama-4-Maverick-17B-128E-Instruct-FP8
diff --git a/metagpt/configs/llm_config.py b/metagpt/configs/llm_config.py
index 2ceef2e2a..f17904d8f 100644
--- a/metagpt/configs/llm_config.py
+++ b/metagpt/configs/llm_config.py
@@ -43,6 +43,7 @@ class LLMType(Enum):
OPENROUTER_REASONING = "openrouter_reasoning"
BEDROCK = "bedrock"
ARK = "ark" # https://www.volcengine.com/docs/82379/1263482#python-sdk
+ LLAMA_API = "llama_api"
def __missing__(self, key):
return self.OPENAI
diff --git a/metagpt/provider/openai_api.py b/metagpt/provider/openai_api.py
index c608d1d29..1263865bb 100644
--- a/metagpt/provider/openai_api.py
+++ b/metagpt/provider/openai_api.py
@@ -52,6 +52,7 @@ from metagpt.utils.token_counter import (
LLMType.DEEPSEEK,
LLMType.SILICONFLOW,
LLMType.OPENROUTER,
+ LLMType.LLAMA_API,
]
)
class OpenAILLM(BaseLLM):
diff --git a/metagpt/utils/token_counter.py b/metagpt/utils/token_counter.py
index 80bdacae5..e8bf9d811 100644
--- a/metagpt/utils/token_counter.py
+++ b/metagpt/utils/token_counter.py
@@ -113,6 +113,10 @@ TOKEN_COSTS = {
"doubao-pro-128k-240515": {"prompt": 0.0007, "completion": 0.0013},
"llama3-70b-llama3-70b-instruct": {"prompt": 0.0, "completion": 0.0},
"llama3-8b-llama3-8b-instruct": {"prompt": 0.0, "completion": 0.0},
+ "llama-4-Scout-17B-16E-Instruct-FP8" : {"prompt": 0.0, "completion": 0.0}, # start, for Llama API
+ "llama-4-Maverick-17B-128E-Instruct-FP8": {"prompt": 0.0, "completion": 0.0},
+ "llama-3.3-8B-Instruct": {"prompt": 0.0, "completion": 0.0},
+ "llama-3.3-70B-Instruct": {"prompt": 0.0, "completion": 0.0}, # end, for Llama API
}