feat(blocks): Add AI/ML API support to LLM blocks

autogpt_platform/backend/.env.example

@@ -63,6 +63,7 @@ GOOGLE_CLIENT_SECRET=
 # LLM
 OPENAI_API_KEY=
 ANTHROPIC_API_KEY=
+AIML_API_KEY=
 GROQ_API_KEY=
 OPEN_ROUTER_API_KEY=
 

autogpt_platform/backend/backend/blocks/llm.py

@@ -35,6 +35,7 @@
     ProviderName.OLLAMA,
     ProviderName.OPENAI,
     ProviderName.OPEN_ROUTER,
+    ProviderName.AIML,
 ]
 AICredentials = CredentialsMetaInput[LLMProviderName, Literal["api_key"]]
 
@@ -98,6 +99,12 @@ class LlmModel(str, Enum, metaclass=LlmModelMeta):
     # Anthropic models
     CLAUDE_3_5_SONNET = "claude-3-5-sonnet-latest"
     CLAUDE_3_HAIKU = "claude-3-haiku-20240307"
+    # Aiml models
+    AIML_QWEN2_5_72B = "Qwen/Qwen2.5-72B-Instruct-Turbo"
+    AIML_LLAMA3_1_70B = "nvidia/llama-3.1-nemotron-70b-instruct"
+    AIML_LLAMA3_3_70B = "meta-llama/Llama-3.3-70B-Instruct-Turbo"
+    AIML_META_LLAMA_3_1_70B = "meta-llama/Meta-Llama-3.1-70B-Instruct-Turbo"
+    AIML_LLAMA_3_2_3B = "meta-llama/Llama-3.2-3B-Instruct-Turbo"
     # Groq models
     LLAMA3_8B = "llama3-8b-8192"
     LLAMA3_70B = "llama3-70b-8192"
@@ -154,6 +161,11 @@ def context_window(self) -> int:
     LlmModel.GPT3_5_TURBO: ModelMetadata("openai", 16385),
     LlmModel.CLAUDE_3_5_SONNET: ModelMetadata("anthropic", 200000),
     LlmModel.CLAUDE_3_HAIKU: ModelMetadata("anthropic", 200000),
+    LlmModel.AIML_QWEN2_5_72B: ModelMetadata("aiml", 32000),
+    LlmModel.AIML_LLAMA3_1_70B: ModelMetadata("aiml", 128000),
+    LlmModel.AIML_LLAMA3_3_70B: ModelMetadata("aiml", 128000),
+    LlmModel.AIML_META_LLAMA_3_1_70B: ModelMetadata("aiml", 131000),
+    LlmModel.AIML_LLAMA_3_2_3B: ModelMetadata("aiml", 128000),
     LlmModel.LLAMA3_8B: ModelMetadata("groq", 8192),
     LlmModel.LLAMA3_70B: ModelMetadata("groq", 8192),
     LlmModel.MIXTRAL_8X7B: ModelMetadata("groq", 32768),
@@ -433,6 +445,23 @@ def llm_call(
                 response.usage.prompt_tokens if response.usage else 0,
                 response.usage.completion_tokens if response.usage else 0,
             )
+        elif provider == "aiml":
+            client = openai.OpenAI(
+                base_url="https://api.aimlapi.com/v2",
+                api_key=credentials.api_key.get_secret_value(),
+            )
+
+            completion = client.chat.completions.create(
+                model=llm_model.value,
+                messages=prompt,  # type: ignore
+                max_tokens=max_tokens,
+            )
+
+            return (
+                completion.choices[0].message.content or "",
+                completion.usage.prompt_tokens if completion.usage else 0,
+                completion.usage.completion_tokens if completion.usage else 0,
+            )
         else:
             raise ValueError(f"Unsupported LLM provider: {provider}")
 
@@ -509,16 +538,19 @@ def parse_response(resp: str) -> tuple[dict[str, Any], str | None]:
                 if input_data.expected_format:
                     parsed_dict, parsed_error = parse_response(response_text)
                     if not parsed_error:
-                        yield "response", {
-                            k: (
-                                json.loads(v)
-                                if isinstance(v, str)
-                                and v.startswith("[")
-                                and v.endswith("]")
-                                else (", ".join(v) if isinstance(v, list) else v)
-                            )
-                            for k, v in parsed_dict.items()
-                        }
+                        yield (
+                            "response",
+                            {
+                                k: (
+                                    json.loads(v)
+                                    if isinstance(v, str)
+                                    and v.startswith("[")
+                                    and v.endswith("]")
+                                    else (", ".join(v) if isinstance(v, list) else v)
+                                )
+                                for k, v in parsed_dict.items()
+                            },
+                        )
                         return
                 else:
                     yield "response", {"response": response_text}

autogpt_platform/backend/backend/data/block_cost_config.py

@@ -20,6 +20,7 @@
 from backend.data.block import Block
 from backend.data.cost import BlockCost, BlockCostType
 from backend.integrations.credentials_store import (
+    aiml_credentials,
     anthropic_credentials,
     did_credentials,
     groq_credentials,
@@ -43,6 +44,11 @@
     LlmModel.GPT3_5_TURBO: 1,
     LlmModel.CLAUDE_3_5_SONNET: 4,
     LlmModel.CLAUDE_3_HAIKU: 1,
+    LlmModel.AIML_QWEN2_5_72B: 1,
+    LlmModel.AIML_LLAMA3_1_70B: 1,
+    LlmModel.AIML_LLAMA3_3_70B: 1,
+    LlmModel.AIML_META_LLAMA_3_1_70B: 1,
+    LlmModel.AIML_LLAMA_3_2_3B: 1,
     LlmModel.LLAMA3_8B: 1,
     LlmModel.LLAMA3_70B: 1,
     LlmModel.MIXTRAL_8X7B: 1,
@@ -142,6 +148,23 @@
         for model, cost in MODEL_COST.items()
         if MODEL_METADATA[model].provider == "open_router"
     ]
+    # AI/ML Api Models
+    + [
+        BlockCost(
+            cost_type=BlockCostType.RUN,
+            cost_filter={
+                "model": model,
+                "credentials": {
+                    "id": aiml_credentials.id,
+                    "provider": aiml_credentials.provider,
+                    "type": aiml_credentials.type,
+                },
+            },
+            cost_amount=cost,
+        )
+        for model, cost in MODEL_COST.items()
+        if MODEL_METADATA[model].provider == "aiml"
+    ]
 )
 
 # =============== This is the exhaustive list of cost for each Block =============== #

autogpt_platform/backend/backend/integrations/credentials_store.py

@@ -49,6 +49,13 @@
     title="Use Credits for OpenAI",
     expires_at=None,
 )
+aiml_credentials = APIKeyCredentials(
+    id="aad82a89-9794-4ebb-977f-d736aa5260a3",
+    provider="aiml",
+    api_key=SecretStr(settings.secrets.aiml_api_key),
+    title="Use Credits for AI/ML",
+    expires_at=None,
+)
 anthropic_credentials = APIKeyCredentials(
     id="24e5d942-d9e3-4798-8151-90143ee55629",
     provider="anthropic",
@@ -98,6 +105,7 @@
     ideogram_credentials,
     replicate_credentials,
     openai_credentials,
+    aiml_credentials,
     anthropic_credentials,
     groq_credentials,
     did_credentials,
@@ -145,6 +153,8 @@ def get_all_creds(self, user_id: str) -> list[Credentials]:
             all_credentials.append(replicate_credentials)
         if settings.secrets.openai_api_key:
             all_credentials.append(openai_credentials)
+        if settings.secrets.aiml_api_key:
+            all_credentials.append(aiml_credentials)
         if settings.secrets.anthropic_api_key:
             all_credentials.append(anthropic_credentials)
         if settings.secrets.did_api_key:

autogpt_platform/backend/backend/integrations/providers.py

@@ -4,6 +4,7 @@
 # --8<-- [start:ProviderName]
 class ProviderName(str, Enum):
     ANTHROPIC = "anthropic"
+    AIML = "aiml"
     COMPASS = "compass"
     DISCORD = "discord"
     D_ID = "d_id"

autogpt_platform/backend/backend/util/settings.py

@@ -266,6 +266,7 @@ class Secrets(UpdateTrackingModel["Secrets"], BaseSettings):
     )
 
     openai_api_key: str = Field(default="", description="OpenAI API key")
+    aiml_api_key: str = Field(default="", description="AI/ML API key")
     anthropic_api_key: str = Field(default="", description="Anthropic API key")
     groq_api_key: str = Field(default="", description="Groq API key")
     open_router_api_key: str = Field(default="", description="Open Router API Key")

autogpt_platform/frontend/src/app/profile/page.tsx

@@ -103,6 +103,7 @@ export default function PrivatePage() {
       "6b9fc200-4726-4973-86c9-cd526f5ce5db", // Replicate
       "53c25cb8-e3ee-465c-a4d1-e75a4c899c2a", // OpenAI
       "24e5d942-d9e3-4798-8151-90143ee55629", // Anthropic
+      "aad82a89-9794-4ebb-977f-d736aa5260a3", // AI/ML
       "4ec22295-8f97-4dd1-b42b-2c6957a02545", // Groq
       "7f7b0654-c36b-4565-8fa7-9a52575dfae2", // D-ID
       "7f26de70-ba0d-494e-ba76-238e65e7b45f", // Jina

autogpt_platform/frontend/src/app/store/(user)/integrations/page.tsx

@@ -103,6 +103,7 @@ export default function PrivatePage() {
       "6b9fc200-4726-4973-86c9-cd526f5ce5db", // Replicate
       "53c25cb8-e3ee-465c-a4d1-e75a4c899c2a", // OpenAI
       "24e5d942-d9e3-4798-8151-90143ee55629", // Anthropic
+      "aad82a89-9794-4ebb-977f-d736aa5260a3", // AI/ML
       "4ec22295-8f97-4dd1-b42b-2c6957a02545", // Groq
       "7f7b0654-c36b-4565-8fa7-9a52575dfae2", // D-ID
       "7f26de70-ba0d-494e-ba76-238e65e7b45f", // Jina

autogpt_platform/frontend/src/components/integrations/credentials-input.tsx

@@ -52,6 +52,7 @@ export const providerIcons: Record<
   github: FaGithub,
   google: FaGoogle,
   groq: fallbackIcon,
+  aiml: fallbackIcon,
   notion: NotionLogoIcon,
   discord: FaDiscord,
   d_id: fallbackIcon,

autogpt_platform/frontend/src/components/integrations/credentials-provider.tsx

@@ -16,6 +16,7 @@ const CREDENTIALS_PROVIDER_NAMES = Object.values(
 
 // --8<-- [start:CredentialsProviderNames]
 const providerDisplayNames: Record<CredentialsProviderName, string> = {
+  aiml: "AI/ML",
   anthropic: "Anthropic",
   discord: "Discord",
   d_id: "D-ID",

autogpt_platform/frontend/src/lib/autogpt-server-api/types.ts

@@ -102,6 +102,7 @@ export type CredentialsType = "api_key" | "oauth2";
 // --8<-- [start:BlockIOCredentialsSubSchema]
 export const PROVIDER_NAMES = {
   ANTHROPIC: "anthropic",
+  AIML: "aiml",
   D_ID: "d_id",
   DISCORD: "discord",
   E2B: "e2b",

classic/forge/forge/llm/providers/aiml.py

@@ -0,0 +1,135 @@
+from __future__ import annotations
+
+import enum
+import logging
+from typing import Any, Optional
+
+import tiktoken
+from pydantic import SecretStr
+
+from forge.models.config import UserConfigurable
+
+from ._openai_base import BaseOpenAIChatProvider
+from .schema import (
+    ChatModelInfo,
+    ModelProviderBudget,
+    ModelProviderConfiguration,
+    ModelProviderCredentials,
+    ModelProviderName,
+    ModelProviderSettings,
+    ModelTokenizer,
+)
+
+
+class AimlModelName(str, enum.Enum):
+    AIML_QWEN2_5_72B = "Qwen/Qwen2.5-72B-Instruct-Turbo"
+    AIML_LLAMA3_1_70B = "nvidia/llama-3.1-nemotron-70b-instruct"
+    AIML_LLAMA3_3_70B = "meta-llama/Llama-3.3-70B-Instruct-Turbo"
+    AIML_LLAMA_3_2_3B = "meta-llama/Llama-3.2-3B-Instruct-Turbo"
+    AIML_META_LLAMA_3_1_70B = "meta-llama/Meta-Llama-3.1-70B-Instruct-Turbo"
+
+
+AIML_CHAT_MODELS = {
+    info.name: info
+    for info in [
+        ChatModelInfo(
+            name=AimlModelName.AIML_QWEN2_5_72B,
+            provider_name=ModelProviderName.AIML,
+            prompt_token_cost=1.26 / 1e6,
+            completion_token_cost=1.26 / 1e6,
+            max_tokens=32000,
+            has_function_call_api=False,
+        ),
+        ChatModelInfo(
+            name=AimlModelName.AIML_LLAMA3_1_70B,
+            provider_name=ModelProviderName.AIML,
+            prompt_token_cost=0.368 / 1e6,
+            completion_token_cost=0.42 / 1e6,
+            max_tokens=128000,
+            has_function_call_api=False,
+        ),
+        ChatModelInfo(
+            name=AimlModelName.AIML_LLAMA3_3_70B,
+            provider_name=ModelProviderName.AIML,
+            prompt_token_cost=0.924 / 1e6,
+            completion_token_cost=0.924 / 1e6,
+            max_tokens=128000,
+            has_function_call_api=False,
+        ),
+        ChatModelInfo(
+            name=AimlModelName.AIML_META_LLAMA_3_1_70B,
+            provider_name=ModelProviderName.AIML,
+            prompt_token_cost=0.063 / 1e6,
+            completion_token_cost=0.063 / 1e6,
+            max_tokens=131000,
+            has_function_call_api=False,
+        ),
+        ChatModelInfo(
+            name=AimlModelName.AIML_LLAMA_3_2_3B,
+            provider_name=ModelProviderName.AIML,
+            prompt_token_cost=0.924 / 1e6,
+            completion_token_cost=0.924 / 1e6,
+            max_tokens=128000,
+            has_function_call_api=False,
+        ),
+    ]
+}
+
+
+class AimlCredentials(ModelProviderCredentials):
+    """Credentials for Aiml."""
+
+    api_key: SecretStr = UserConfigurable(from_env="AIML_API_KEY")  # type: ignore
+    api_base: Optional[SecretStr] = UserConfigurable(
+        default=None, from_env="AIML_API_BASE_URL"
+    )
+
+    def get_api_access_kwargs(self) -> dict[str, str]:
+        return {
+            k: v.get_secret_value()
+            for k, v in {
+                "api_key": self.api_key,
+                "base_url": self.api_base,
+            }.items()
+            if v is not None
+        }
+
+
+class AimlSettings(ModelProviderSettings):
+    credentials: Optional[AimlCredentials]  # type: ignore
+    budget: ModelProviderBudget  # type: ignore
+
+
+class AimlProvider(BaseOpenAIChatProvider[AimlModelName, AimlSettings]):
+    CHAT_MODELS = AIML_CHAT_MODELS
+    MODELS = CHAT_MODELS
+
+    default_settings = AimlSettings(
+        name="aiml_provider",
+        description="Provides access to AIML's API.",
+        configuration=ModelProviderConfiguration(),
+        credentials=None,
+        budget=ModelProviderBudget(),
+    )
+
+    _settings: AimlSettings
+    _configuration: ModelProviderConfiguration
+    _credentials: AimlCredentials
+    _budget: ModelProviderBudget
+
+    def __init__(
+        self,
+        settings: Optional[AimlSettings] = None,
+        logger: Optional[logging.Logger] = None,
+    ):
+        super(AimlProvider, self).__init__(settings=settings, logger=logger)
+
+        from openai import AsyncOpenAI
+
+        self._client = AsyncOpenAI(
+            **self._credentials.get_api_access_kwargs()  # type: ignore
+        )
+
+    def get_tokenizer(self, model_name: AimlModelName) -> ModelTokenizer[Any]:
+        # HACK: No official tokenizer is available for AIML
+        return tiktoken.encoding_for_model("gpt-3.5-turbo")