letta-nightly 0.10.0.dev20250801104513__py3-none-any.whl → 0.10.0.dev20250802104311__py3-none-any.whl

letta/schemas/providers/ollama.py

@@ -1,7 +1,6 @@
 from typing import Literal
 
 import aiohttp
-import requests
 from pydantic import Field
 
 from letta.constants import DEFAULT_EMBEDDING_CHUNK_SIZE
@@ -43,7 +42,7 @@ class OllamaProvider(OpenAIProvider):
 
         configs = []
         for model in response_json["models"]:
-            context_window = self.get_model_context_window(model["name"])
+            context_window = await self._get_model_context_window(model["name"])
             if context_window is None:
                 print(f"Ollama model {model['name']} has no context window, using default 32000")
                 context_window = 32000
@@ -75,7 +74,7 @@ class OllamaProvider(OpenAIProvider):
 
         configs = []
         for model in response_json["models"]:
-            embedding_dim = await self._get_model_embedding_dim_async(model["name"])
+            embedding_dim = await self._get_model_embedding_dim(model["name"])
             if not embedding_dim:
                 print(f"Ollama model {model['name']} has no embedding dimension, using default 1024")
                 # continue
@@ -92,63 +91,50 @@ class OllamaProvider(OpenAIProvider):
             )
         return configs
 
-    def get_model_context_window(self, model_name: str) -> int | None:
-        """Gets model context window for Ollama. As this can look different based on models,
-        we use the following for guidance:
-
-        "llama.context_length": 8192,
-        "llama.embedding_length": 4096,
-        source: https://github.com/ollama/ollama/blob/main/docs/api.md#show-model-information
-
-        FROM 2024-10-08
-        Notes from vLLM around keys
-        source: https://github.com/vllm-project/vllm/blob/72ad2735823e23b4e1cc79b7c73c3a5f3c093ab0/vllm/config.py#L3488
-
-        possible_keys = [
-            # OPT
-            "max_position_embeddings",
-            # GPT-2
-            "n_positions",
-            # MPT
-            "max_seq_len",
-            # ChatGLM2
-            "seq_length",
-            # Command-R
-            "model_max_length",
-            # Whisper
-            "max_target_positions",
-            # Others
-            "max_sequence_length",
-            "max_seq_length",
-            "seq_len",
-        ]
-        max_position_embeddings
-        parse model cards: nous, dolphon, llama
-        """
+    async def _get_model_context_window(self, model_name: str) -> int | None:
         endpoint = f"{self.base_url}/api/show"
-        payload = {"name": model_name, "verbose": True}
-        response = requests.post(endpoint, json=payload)
-        if response.status_code != 200:
-            return None
+        payload = {"name": model_name}
 
         try:
-            model_info = response.json()
-            # Try to extract context window from model parameters
-            if "model_info" in model_info and "llama.context_length" in model_info["model_info"]:
-                return int(model_info["model_info"]["llama.context_length"])
-        except Exception:
-            pass
-        logger.warning(f"Failed to get model context window for {model_name}")
+            async with aiohttp.ClientSession() as session:
+                async with session.post(endpoint, json=payload) as response:
+                    if response.status != 200:
+                        error_text = await response.text()
+                        logger.warning(f"Failed to get model info for {model_name}: {response.status} - {error_text}")
+                        return None
+
+                    response_json = await response.json()
+                    model_info = response_json.get("model_info", {})
+
+                    if architecture := model_info.get("general.architecture"):
+                        if context_length := model_info.get(f"{architecture}.context_length"):
+                            return int(context_length)
+
+        except Exception as e:
+            logger.warning(f"Failed to get model context window for {model_name} with error: {e}")
+
         return None
 
-    async def _get_model_embedding_dim_async(self, model_name: str):
-        async with aiohttp.ClientSession() as session:
-            async with session.post(f"{self.base_url}/api/show", json={"name": model_name, "verbose": True}) as response:
-                response_json = await response.json()
+    async def _get_model_embedding_dim(self, model_name: str) -> int | None:
+        endpoint = f"{self.base_url}/api/show"
+        payload = {"name": model_name}
+
+        try:
+            async with aiohttp.ClientSession() as session:
+                async with session.post(endpoint, json=payload) as response:
+                    if response.status != 200:
+                        error_text = await response.text()
+                        logger.warning(f"Failed to get model info for {model_name}: {response.status} - {error_text}")
+                        return None
 
-        if "model_info" not in response_json:
-            if "error" in response_json:
-                logger.warning("Ollama fetch model info error for %s: %s", model_name, response_json["error"])
-            return None
+                    response_json = await response.json()
+                    model_info = response_json.get("model_info", {})
 
-        return response_json["model_info"].get("embedding_length")
+                    if architecture := model_info.get("general.architecture"):
+                        if embedding_length := model_info.get(f"{architecture}.embedding_length"):
+                            return int(embedding_length)
+
+        except Exception as e:
+            logger.warning(f"Failed to get model embedding dimension for {model_name} with error: {e}")
+
+        return None

{letta_nightly-0.10.0.dev20250801104513.dist-info → letta_nightly-0.10.0.dev20250802104311.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: letta-nightly
-Version: 0.10.0.dev20250801104513
+Version: 0.10.0.dev20250802104311
 Summary: Create LLM agents with long-term memory and custom tools
 License: Apache License
 Author: Letta Team

{letta_nightly-0.10.0.dev20250801104513.dist-info → letta_nightly-0.10.0.dev20250802104311.dist-info}/RECORD

@@ -287,7 +287,7 @@ letta/schemas/providers/groq.py,sha256=AquJQH-Y5-s75Nj2_X7xavuWUu5F2bSvHjAZ1Gfpe
 letta/schemas/providers/letta.py,sha256=DD42qaJoR7YvaWIB0Y1oMFGQgcI3BDe0X6--5DB1px4,1585
 letta/schemas/providers/lmstudio.py,sha256=fx1lfLG4K1x6RUeHEXr9pMnQ1IgMpK1XmW_Y23riwgw,4303
 letta/schemas/providers/mistral.py,sha256=EjFF6YcfN5jBjCfnZw3ECv_3qYuG0HVb7B0VoYk-jKU,1866
-letta/schemas/providers/ollama.py,sha256=IMnlVUrwvJeuIMAgD6buJq5Ccows2R4KZzFJ9QF8VNU,6452
+letta/schemas/providers/ollama.py,sha256=iVx9xxrulG9ohbhk4kMtfryCnMgmynWg_NL4SesGX5U,6253
 letta/schemas/providers/openai.py,sha256=Et2NoOPWQ4xIn_WMP2ingZM_heLFM7SCtY6kHxCi9Tw,11042
 letta/schemas/providers/together.py,sha256=2zFca6Jy08r1ANrdvtlSIduyDr8ek9Tt1yYiz1S-5g8,3422
 letta/schemas/providers/vllm.py,sha256=OK98JyUXsI5Ygr1xSG9WDH6JqQXDifVSccdGwRW_gUQ,2482
@@ -451,8 +451,8 @@ letta/templates/summary_request_text.j2,sha256=ZttQwXonW2lk4pJLYzLK0pmo4EO4EtUUI
 letta/templates/template_helper.py,sha256=CzkLA-n_BCyvguP4c0V-O_YjQY1vl059MAQwbjt198s,1364
 letta/types/__init__.py,sha256=hokKjCVFGEfR7SLMrtZsRsBfsC7yTIbgKPLdGg4K1eY,147
 letta/utils.py,sha256=Fwwe2imHRamc_kucAATo8NXhwDG5NBoOIYmBaERXUhM,38384
-letta_nightly-0.10.0.dev20250801104513.dist-info/LICENSE,sha256=mExtuZ_GYJgDEI38GWdiEYZizZS4KkVt2SF1g_GPNhI,10759
-letta_nightly-0.10.0.dev20250801104513.dist-info/METADATA,sha256=M1XrCufQlCD86lBxCDTFfzhSgSwzjjVv2bWm0dfOv8s,23310
-letta_nightly-0.10.0.dev20250801104513.dist-info/WHEEL,sha256=FMvqSimYX_P7y0a7UY-_Mc83r5zkBZsCYPm7Lr0Bsq4,88
-letta_nightly-0.10.0.dev20250801104513.dist-info/entry_points.txt,sha256=2zdiyGNEZGV5oYBuS-y2nAAgjDgcC9yM_mHJBFSRt5U,40
-letta_nightly-0.10.0.dev20250801104513.dist-info/RECORD,,
+letta_nightly-0.10.0.dev20250802104311.dist-info/LICENSE,sha256=mExtuZ_GYJgDEI38GWdiEYZizZS4KkVt2SF1g_GPNhI,10759
+letta_nightly-0.10.0.dev20250802104311.dist-info/METADATA,sha256=P1uhyLugyzH1yfVbh-7l6wxYlfRYK8ZK5svXSpWkzzE,23310
+letta_nightly-0.10.0.dev20250802104311.dist-info/WHEEL,sha256=FMvqSimYX_P7y0a7UY-_Mc83r5zkBZsCYPm7Lr0Bsq4,88
+letta_nightly-0.10.0.dev20250802104311.dist-info/entry_points.txt,sha256=2zdiyGNEZGV5oYBuS-y2nAAgjDgcC9yM_mHJBFSRt5U,40
+letta_nightly-0.10.0.dev20250802104311.dist-info/RECORD,,