docs(aimlapi): document aimlapi.com as an LLM/embedding provider

.claude/skills/cognee-integrations/SKILL.md

@@ -1,6 +1,6 @@
 ---
 name: cognee-integrations
-description: Use when the user wants to connect cognee to external services — switching LLM or embedding providers (OpenAI, Azure, Gemini, Anthropic, Ollama, OpenRouter), changing databases (Postgres, PGVector, Neo4j, Neptune, Turso), S3 storage, or the MCP server for IDE integration.
+description: Use when the user wants to connect cognee to external services — switching LLM or embedding providers (OpenAI, Azure, Gemini, Anthropic, Ollama, OpenRouter, aimlapi.com), changing databases (Postgres, PGVector, Neo4j, Neptune, Turso), S3 storage, or the MCP server for IDE integration.
 ---
 
 # Set up cognee integrations
@@ -16,6 +16,7 @@ Default is OpenAI (`LLM_API_KEY` is all you need). To switch, set
 `LLM_PROVIDER`, `LLM_MODEL`, `LLM_API_KEY`, and (where relevant)
 `LLM_ENDPOINT` / `LLM_API_VERSION`:
 
+- **aimlapi.com**: `LLM_PROVIDER=custom`, `LLM_MODEL=aiml/<catalogue-id>`, `LLM_ENDPOINT=https://api.aimlapi.com/v1`. Embeddings drop the `aiml/` prefix (litellm has no aiml embedding route) and rely on `EMBEDDING_ENDPOINT` instead.
 - **Azure OpenAI**: `LLM_PROVIDER=azure`, `LLM_MODEL=azure/gpt-4o-mini`, endpoint + api version required.
 - **Gemini** (no extra needed): `LLM_PROVIDER=gemini`, `LLM_MODEL=gemini/gemini-2.0-flash-exp`.
 - **Anthropic** (`cognee[anthropic]`): `LLM_PROVIDER=anthropic`, model e.g. `claude-3-5-sonnet-20241022`.

.env.template

@@ -781,6 +781,28 @@ WEB_SCRAPER_MAX_DELAY=10.0
 # Uncomment + fill values to switch providers.
 ###############################################################################
 
+########## aimlapi.com ########################################################
+# OpenAI-compatible gateway; litellm routes it under the "aiml/" prefix, so no
+# extra install is needed. LLM_PROVIDER must be "custom" -- an "aiml/" prefix on
+# its own raises ProviderNotDeducibleError, exactly like "openrouter/".
+# Catalogue: https://api.aimlapi.com/v1/models?include=all
+# Note: litellm's own env var for this provider is AIML_API_KEY, but cognee
+# never reads it -- cognee passes api_key to litellm explicitly, so the key
+# must be in LLM_API_KEY (setting only AIML_API_KEY raises LLMAPIKeyNotSetError).
+#LLM_API_KEY="<<go-get-one-yourself"
+#LLM_PROVIDER="custom"
+#LLM_MODEL="aiml/openai/gpt-4o-mini"
+#LLM_ENDPOINT="https://api.aimlapi.com/v1"
+# Embeddings are configured independently. litellm has no "aiml" embedding
+# route, so the embedding model carries NO "aiml/" prefix -- EMBEDDING_ENDPOINT
+# is what points the OpenAI-compatible route at aimlapi.com.
+#EMBEDDING_PROVIDER="custom"
+#EMBEDDING_API_KEY="<<go-get-one-yourself"
+#EMBEDDING_MODEL="openai/text-embedding-3-large"
+#EMBEDDING_ENDPOINT="https://api.aimlapi.com/v1"
+#EMBEDDING_DIMENSIONS=3072
+#EMBEDDING_MAX_COMPLETION_TOKENS=8191
+
 ########## Azure OpenAI (API key auth) ########################################
 #LLM_PROVIDER="azure"
 #LLM_MODEL="azure/gpt-5-mini"

CLAUDE.md

@@ -415,7 +415,21 @@ Four flags trade memory features for speed. Know what each turns off before flip
 
 ### LLM Provider Configuration
 
-Supported providers: OpenAI (default), Azure OpenAI, Google Gemini, Anthropic, AWS Bedrock, Ollama, LM Studio, Custom (OpenAI-compatible APIs)
+Supported providers: aimlapi.com, OpenAI (default), Azure OpenAI, Google Gemini, Anthropic, AWS Bedrock, Ollama, LM Studio, Custom (OpenAI-compatible APIs)
+
+#### aimlapi.com
+```bash
+LLM_PROVIDER="custom"
+LLM_MODEL="aiml/openai/gpt-4o-mini"
+LLM_ENDPOINT="https://api.aimlapi.com/v1"
+LLM_API_KEY="your_aimlapi_key"  # not AIML_API_KEY: cognee passes the key to litellm itself
+# Embeddings are separate; no "aiml/" prefix (litellm has no aiml embedding route)
+EMBEDDING_PROVIDER="custom"
+EMBEDDING_MODEL="openai/text-embedding-3-large"
+EMBEDDING_ENDPOINT="https://api.aimlapi.com/v1"
+EMBEDDING_API_KEY="your_aimlapi_key"
+EMBEDDING_DIMENSIONS=3072
+```
 
 #### OpenAI (Recommended - Minimal Setup)
 ```bash

catalog/entries/integrations/aimlapi.yaml

@@ -0,0 +1,27 @@
+id: aimlapi
+title: aimlapi.com as the cognee LLM provider
+kind: integration
+stack: llm-provider
+tags:
+  - aimlapi
+  - llm-provider
+  - openai-compatible
+  - litellm
+  - embeddings
+summary: Run cognee's LLM and embedding calls on aimlapi.com, an OpenAI-compatible gateway, using litellm's built-in "aiml/" route and no extra install.
+what_youll_build: A cognee pipeline that builds a knowledge graph and answers a recall query with both extraction and embeddings served by aimlapi.com.
+quickstart: |
+  git clone https://github.com/topoteretes/cognee.git
+  cd cognee
+  uv pip install cognee
+  export LLM_API_KEY=your_aimlapi_key
+  python examples/guides/aimlapi_example.py
+expected_output: |
+  A "Knowledge graph built successfully" line followed by a recall answer
+  synthesized from the sample text. The LLM call is routed through litellm's
+  aiml/ prefix and the embedding call through the aimlapi.com endpoint, so no
+  OpenAI key is involved.
+difficulty: easy
+repo: topoteretes/cognee
+path: examples/guides/aimlapi_example.py
+example_path: examples/guides/aimlapi_example.py

examples/guides/aimlapi_example.py

@@ -0,0 +1,97 @@
+"""Example: Running Cognee on aimlapi.com.
+
+aimlapi.com is an OpenAI-compatible gateway serving hundreds of chat models
+behind one key. Cognee reaches it through litellm, so no adapter is needed:
+set ``LLM_PROVIDER="custom"`` and prefix the catalogue id with ``aiml/``.
+
+- LLM Provider: aimlapi.com (``aiml/openai/gpt-4o-mini``)
+- Embeddings: aimlapi.com (``openai/text-embedding-3-large``)
+- Local embedded database stack (Kuzu, LanceDB, SQLite)
+
+Requires an aimlapi.com key exported as ``LLM_API_KEY``. Get one at
+https://aimlapi.com/app/keys — the model catalogue is public at
+https://api.aimlapi.com/v1/models?include=all
+"""
+
+import os
+import asyncio
+import tempfile
+from pathlib import Path
+
+# Setup temp directory to keep this example self-contained
+_DATA_DIR = tempfile.mkdtemp(prefix="cognee_aimlapi_example_")
+os.environ["ENABLE_BACKEND_ACCESS_CONTROL"] = "false"
+os.environ["CACHING"] = "false"
+
+# Configure aimlapi.com environment settings.
+# LLM_PROVIDER must be "custom": cognee's provider enum is deliberately small and
+# an "aiml/" prefix alone raises ProviderNotDeducibleError, exactly like "openrouter/".
+os.environ["LLM_PROVIDER"] = "custom"
+os.environ["LLM_MODEL"] = "aiml/openai/gpt-4o-mini"
+os.environ["LLM_ENDPOINT"] = "https://api.aimlapi.com/v1"
+os.environ.setdefault("LLM_API_KEY", "")  # export your key before running
+
+# Embeddings are configured independently of the LLM. litellm has no "aiml"
+# embedding route, so the embedding model carries NO "aiml/" prefix -- it is the
+# endpoint that points the OpenAI-compatible route at aimlapi.com.
+os.environ["EMBEDDING_PROVIDER"] = "custom"
+os.environ["EMBEDDING_MODEL"] = "openai/text-embedding-3-large"
+os.environ["EMBEDDING_ENDPOINT"] = "https://api.aimlapi.com/v1"
+os.environ["EMBEDDING_API_KEY"] = os.environ["LLM_API_KEY"]
+os.environ["EMBEDDING_DIMENSIONS"] = "3072"
+os.environ["EMBEDDING_MAX_COMPLETION_TOKENS"] = "8191"
+
+import cognee  # noqa: E402
+from cognee.modules.search.types import SearchType  # noqa: E402
+from cognee.infrastructure.llm.config import get_llm_config  # noqa: E402
+
+# Force local embedded stack configuration
+cognee.config.set_graph_database_provider("kuzu")
+cognee.config.set_vector_db_provider("lancedb")
+cognee.config.data_root_directory(str(Path(_DATA_DIR) / "data"))
+cognee.config.system_root_directory(str(Path(_DATA_DIR) / "system"))
+
+
+SAMPLE_TEXT = """\
+Cognee is an open-source library that helps developers turn documents into AI memory.
+It builds semantic graphs, indexes entities, and stores vectors to enable structured retrieval.
+Cognee supports hosted gateways such as aimlapi.com as well as local execution via Ollama.
+"""
+
+
+def banner(title: str) -> None:
+    print("\n" + "=" * 78)
+    print(title)
+    print("=" * 78)
+
+
+async def main() -> None:
+    if not os.environ.get("LLM_API_KEY"):
+        raise SystemExit("Export LLM_API_KEY with your aimlapi.com key before running.")
+
+    # Start from a clean slate in isolated directory
+    await cognee.prune.prune_data()
+    await cognee.prune.prune_system(metadata=True)
+
+    banner("REMEMBER USING AIMLAPI.COM")
+    llm_config = get_llm_config()
+    print(f"Using LLM: {llm_config.llm_model}")
+    print(f"Using Embeddings: {os.environ.get('EMBEDDING_MODEL')}")
+
+    await cognee.remember(SAMPLE_TEXT, dataset_name="aimlapi_demo", self_improvement=False)
+    print("Knowledge graph built successfully.")
+
+    banner("RECALL")
+    query = "What does Cognee help developers do?"
+    results = await cognee.recall(
+        query_text=query,
+        query_type=SearchType.GRAPH_COMPLETION,
+        datasets=["aimlapi_demo"],
+    )
+    print(f"Query: {query}")
+    print("Recall Results:")
+    print(results[0].text if results else "<no results>")
+
+
+if __name__ == "__main__":
+    asyncio.run(main())