letta-ai/letta 0.6.5

v0.6.5

on GitHub

This release includes bug fixes for Ollama embeddings, improved error handling for streaming, improvements in debugging tool callings and agent creation.

⚙️ Providing tool call logs (stdout and stderr)

We now pass back the logs from tool execution in the FunctionResponse object. You can view these logs in the ADE in addition to the tool response:

👾 Simplification of agent creation API

We now allow for simply specifying the LLM and embedding configuration of an agent by specifying the model in the format <provider>/<model>:

curl --request POST \
  --url http://localhost:8283/v1/agents/ \
  --header 'Content-Type: application/json' \
  --data '{
     "memory_blocks": [
    {
      "value": "Name: Sarah",
      "label": "human"
    }, 
    {
      "value": "I am a helpful assistant",
      "label": "persona"
    }
  ],
  "llm": "anthropic/claude-3-5-sonnet-20241022",
  "embedding": "openai/text-embedding-ada-002"
}'

You can also provide the context window, and specify the character limit of memory blocks (e.g. human/persona):

curl --request POST \
  --url http://localhost:8283/v1/agents/ \
  --header 'Content-Type: application/json' \
  --data '{
     "memory_blocks": [
    {
      "value": "Name: Sarah",
      "limit": 5000,
      "label": "human"
    }, 
    {
      "value": "I am a helpful assistant",
      "label": "persona"
    }
  ],
  "llm": "anthropic/claude-3-5-sonnet-20241022",
  "embedding": "openai/text-embedding-ada-002",
  "context_window_limit": 15000
}'

What's Changed

• feat: Catch composio errors and return informative errors in the endpoint by @mattzh72 in #2247
• feat: Rewrite agents by @mattzh72 in #2232
• Fix: Move orm metadata out of write-only Agent pydantic objects by @mattzh72 in #2249
• fix: Changing e2b template id as an env var reflects immediately by @mattzh72 in #2250
• fix: Remove created_by_id check by @mattzh72 in #2252
• chore: Delete metadata.py by @mattzh72 in #2253
• docs: fix typo in prompt by @keenranger in #2256
• chore: Catch orm specific issues in agents v1 routes and simplify tool add/remove from agent by @mattzh72 in #2259
• feat: separate Passages tables by @mlong93 in #2245
• fix: add error logging on stream fails by @cpacker in #2261
• docs: Feature/update docs by @keithf123r in #2257
• fix: Make embedding_endpoint_type an enum by @mattzh72 in #2251
• fix: Fix update agent by @mattzh72 in #2265
• chore: Add comprehensive unit test for adding/removing tools via update agent by @mattzh72 in #2267
• fix: Scale up database by @mattzh72 in #2263
• feat: Adding init tool rule for Anthropic endpoint by @mlong93 in #2262
• fix: fix ollama embeddings by @sarahwooders in #2264
• Create multiplatform docker images by @AliSayyah in #2254
• feat: Add optional llm and embedding handle args to CreateAgent request by @carenthomas in #2260
• fix: refactor sandbox run logic to add status field by @carenthomas in #2248
• chore: Add testing around base tools by @mattzh72 in #2268
• fix: Deprecate in memory function stores in agent.py by @mattzh72 in #2271
• fix: Allow ChildToolRule to work without support for structured outputs by @mlong93 in #2270
• chore: bump version to 0.6.5 by @sarahwooders in #2275

New Contributors

• @keenranger made their first contribution in #2256
• @keithf123r made their first contribution in #2257

Full Changelog: 0.6.4...0.6.5