The AI Toolchain — issue -384, November 30, 2023

Zed v0.113.0 adds GPT-4 Turbo support and updates the default OpenAI model in the assistant panel.

$ git clone --branch v0.113.0 https://github.com/zed-industries/zed.git
# already have the repo? check out this version:
$ git checkout v0.113.0

"assistant": {
    "default_open_ai_model": "gpt-4-0613"
}

• ›Updates assistant.default_open_ai_model in settings.json to default to gpt-4-1106-preview; override with gpt-3.5-turbo-0613, gpt-4-0613, or gpt-4-1106-preview.
• ›Adds support for the gpt-4-1106-preview model in the assistant panel.

Zed v0.112.3 adds seed_search_query_from_cursor setting to control automatic search query population.

$ git clone --branch v0.112.3 https://github.com/zed-industries/zed.git
# already have the repo? check out this version:
$ git checkout v0.112.3

{
  "seed_search_query_from_cursor": "selection"
}

• ›Adds seed_search_query_from_cursor to ~/.zed/settings.json to control whether buffer and project search queries are auto-populated from the cursor; supports values 'always' (default), 'selection' (only when text is selected), and 'never'.

Zed v0.110.2 adds Tailwind autocomplete to six new file types, a channel notification panel, @-mentions, and new Vim mode commands.

$ git clone --branch v0.110.2 https://github.com/zed-industries/zed.git
# already have the repo? check out this version:
$ git checkout v0.110.2

• ›Adds Toggle Vim Mode command to switch Vim emulation on or off from the command palette.
• ›Adds Tailwind CSS autocomplete support to Svelte, Phoenix HEEX, Phoenix ~H sigil in Elixir, ERB, PHP, and Laravel Blade files.
• ›Adds a notification panel in channels that surfaces contact requests and channel invitations.
• ›Adds @-mention support in channel chat, triggering notifications for the mentioned user.
• ›Adds a guest role to channels, set as the default when a new user joins a public channel.

CrewAI v0.1.0 debuts a Python framework for orchestrating collaborative, role-based autonomous AI agents.

$ git clone --branch v0.1.0 https://github.com/crewAIInc/crewAI.git
# already have the repo? check out this version:
$ git checkout v0.1.0

from crewai import Agent, Task, Crew, Process

researcher = Agent(
    role='Researcher',
    goal='Find key facts about quantum computing',
    tools=[search_tool]
)

writer = Agent(
    role='Writer',
    goal='Summarize research into a short briefing',
    tools=[]
)

research_task = Task(description='Research recent quantum computing breakthroughs', tools=[search_tool])
write_task    = Task(description='Write a 200-word executive summary', tools=[])

crew = Crew(
    agents=[researcher, writer],
    tasks=[research_task, write_task],
    process=Process.sequential
)

result = crew.kickoff()
print(result)

• ›Introduces the Agent class for defining autonomous agents with specific roles, goals, and assigned tools.
• ›Introduces the Task class for dynamically creating and assigning tasks with per-task tool specifications.
• ›Introduces the Crew class for grouping agents and coordinating collaborative workflows.
• ›Introduces the Process class supporting sequential task execution for organized, predictable agent pipelines.
• ›Enables inter-agent delegation, allowing agents to autonomously redistribute subtasks among team members at runtime.

Haystack v1.22.0 adds async Pipeline support, new Haystack 2.0 preview components, and expanded model/hardware compatibility.

$ git clone --branch v1.22.0 https://github.com/deepset-ai/haystack.git
# already have the repo? check out this version:
$ git checkout v1.22.0

with open('pipeline.yaml', 'w') as f:
    pipeline.dump(f)

retriever = WebRetriever(search_engine_kwargs={'engine': '<your-engine-id>'})

• ›Adds ByteStream type (with mime_type field) for passing binary raw data across pipeline components in Haystack 2.0.
• ›Adds ChatMessage dataclass to PromptBuilder for structured chat LLM message handling in Haystack 2.0.
• ›Adds AzureOCRDocumentConverter to convert documents via Azure's Document Intelligence Service in Haystack 2.0.
• ›Adds HTMLToDocument component to convert HTML to a Document in Haystack 2.0.
• ›Adds TransformersSimilarityRanker component (renamed from SimilarityRanker) that ranks Document lists by query similarity in Haystack 2.0.

LangChain v0.0.342 adds Databricks Vector Search, Infinity embeddings, agent streaming, and an Amazon Bedrock Knowledge Bases retriever.

$ git clone --branch v0.0.342 https://github.com/langchain-ai/langchain.git
# already have the repo? check out this version:
$ git checkout v0.0.342

from langchain.agents import AgentExecutor

agent_executor = AgentExecutor(agent=agent, tools=tools)

for chunk in agent_executor.stream({"input": "What is the weather in SF?"}):
    print(chunk)

• ›Adds stream() and astream() methods to agents, enabling real-time token-by-token output from agent runs.
• ›Adds RunnableLambda automatic async promotion: when no afunc is provided, an async instance is automatically created from func.
• ›Tracks RunnableAssign as a separate run trace for finer-grained observability in LangSmith.
• ›Adds retriever for Knowledge Bases for Amazon Bedrock, enabling RAG over managed Bedrock knowledge bases.
• ›Adds Databricks Vector Search as a new vector store integration.

LangChain v0.0.341 adds Astra DB chat history and LLM caching, OneNote loader, Outline retriever, and skeleton-of-thought support.

$ git clone --branch v0.0.341 https://github.com/langchain-ai/langchain.git
# already have the repo? check out this version:
$ git checkout v0.0.341

• ›Adds option to prefix config keys in configurable_alts, enabling namespaced configuration for alternative runnables.
• ›New AstraDBChatMessageHistory integration for storing chat message history in Astra DB.
• ›New Astra DB LLM cache classes supporting both exact-match and semantic caching backends.
• ›Adds title metadata field to GoogleDriveLoader when using optional File Loaders.
• ›New OneNote document loader for ingesting Microsoft OneNote content.

LangChain v0.0.339rc3 adds Astra DB chat history and LLM caching, plus title metadata for GoogleDriveLoader.

$ git clone --branch v0.0.339rc3 https://github.com/langchain-ai/langchain.git
# already have the repo? check out this version:
$ git checkout v0.0.339rc3

• ›Adds AstraDBChatMessageHistory integration for storing and retrieving chat message history in Astra DB.
• ›Adds Astra DB LLM cache classes supporting both exact-match and semantic caching backends.
• ›Adds title metadata field to GoogleDriveLoader when using optional File Loaders.

LangChain v0.0.340 adds batch_size to LLM callbacks, partial_variables to prompt templates, and a gpt-crawler template.

$ git clone --branch v0.0.340 https://github.com/langchain-ai/langchain.git
# already have the repo? check out this version:
$ git checkout v0.0.340

from langchain.prompts import HumanMessagePromptTemplate

template = HumanMessagePromptTemplate.from_template(
    "You are a {role}. Answer the following: {question}",
    partial_variables={"role": "cybersecurity analyst"}
)
message = template.format(question="What are common SQL injection patterns?")

• ›Adds batch_size kwarg to the llm_start callback, enabling downstream handlers to know how many inputs are being processed in a single LLM call.
• ›Adds partial_variables support to BaseStringMessagePromptTemplate.from_template(...), allowing partial variable binding directly at template construction.
• ›Adds embed_general_texts method to VoyageEmbeddings for broader embedding coverage.
• ›Adds a new gpt-crawler project template for building RAG pipelines from crawled web content.

LangChain v0.0.339rc0 adds a gpt-crawler template, error rate tracking, and a langchain-core dependency.

$ git clone --branch v0.0.339rc0 https://github.com/langchain-ai/langchain.git
# already have the repo? check out this version:
$ git checkout v0.0.339rc0

• ›Adds a new template for gpt-crawler to enable RAG pipelines over crawled web content.
• ›Adds error rate metric tracking via a new evaluation addition.
• ›Introduces langchain-core as an explicit dependency, extracting core utilities into a dedicated package.

LangChain v0.0.339 adds an Embedchain retriever, llama2-13b-chat-v1 support in BedrockChat, ERNIE-Bot-4 function calling, and search_kwargs for BingSearchAPIWrapper.

$ git clone --branch v0.0.339 https://github.com/langchain-ai/langchain.git
# already have the repo? check out this version:
$ git checkout v0.0.339

from langchain.utilities import BingSearchAPIWrapper

search = BingSearchAPIWrapper(
    search_kwargs={"mkt": "en-US", "count": 5}
)
results = search.run("latest CVE disclosures")

from langchain.chat_models import BedrockChat

llm = BedrockChat(model_id="meta.llama2-13b-chat-v1", region_name="us-east-1")
response = llm.predict("Summarize the OWASP Top 10 for 2023.")

• ›Adds search_kwargs parameter to BingSearchAPIWrapper for passing custom parameters to Bing Search API calls.
• ›Adds llama2-13b-chat-v1 model support to chat_models.BedrockChat.
• ›Adds ERNIE-Bot-4 function calling support.
• ›Adds new Embedchain retriever integration.
• ›Adds YoutubeLoader on-demand language translation support.

LangChain v0.0.338 adds a generic LLM-to-chat-model wrapper, new OctoAI endpoint support, and Neptune graph updates.

$ git clone --branch v0.0.338 https://github.com/langchain-ai/langchain.git
# already have the repo? check out this version:
$ git checkout v0.0.338

• ›Adds a generic LLM wrapper that exposes the chat model interface with a configurable chat prompt format, enabling chat-style interactions through standard LLM backends.
• ›Adds support for new OctoAI endpoints, expanding hosted model coverage.
• ›Updates Neptune graph integration with new capabilities.
• ›Adds execution time tracking to runs.

LangChain v0.0.337 adds RunnableWithMessageHistory, multi-index templates, and input_type for VoyageEmbeddings

$ git clone --branch v0.0.337 https://github.com/langchain-ai/langchain.git
# already have the repo? check out this version:
$ git checkout v0.0.337

from langchain.runnables.history import RunnableWithMessageHistory

chain_with_history = RunnableWithMessageHistory(
    chain,
    get_session_history=get_session_history,
    input_messages_key="input",
    history_messages_key="history",
)
chain_with_history.invoke(
    {"input": "What is LangChain?"},
    config={"configurable": {"session_id": "user-123"}},
)

from langchain.embeddings import VoyageEmbeddings

embeddings = VoyageEmbeddings(
    model="voyage-01",
    input_type="query",
)
result = embeddings.embed_query("What is retrieval-augmented generation?")

• ›Adds input_type field to VoyageEmbeddings for specifying embedding input type.
• ›Adds serialization arguments to Bedrock and ChatBedrock integrations.
• ›Adds optional constructor arguments to FalkorDBGraph for more flexible graph initialization.
• ›Adds ahandle_event to the _all_ callback set, enabling async event handling across all callback types.
• ›Adds RunnableWithMessageHistory, enabling stateful message history management in LCEL chains.

LangChain v0.0.336 adds OAI Assistants with callbacks, limit_to_domains for APIChain, Bedrock Cohere embeddings, Yi model support, and Azure OpenAI v1 completions.

$ git clone --branch v0.0.336 https://github.com/langchain-ai/langchain.git
# already have the repo? check out this version:
$ git checkout v0.0.336

from langchain.chains import APIChain

chain = APIChain.from_llm_and_api_docs(
    llm=llm,
    api_docs=my_api_docs,
    limit_to_domains=["api.example.com", "data.example.org"]
)

from langchain.llms import Ollama

llm = Ollama(
    model="llama2",
    system="You are a concise cybersecurity assistant.",
    template="### Instruction:\n{prompt}\n### Response:"
)

• ›Adds limit_to_domains parameter to APIChain-based tools to restrict which domains the chain is permitted to call.
• ›Adds system prompt and template fields to the Ollama integration, enabling structured prompt control.
• ›Adds model parameter to the DALL-E integration, allowing explicit model selection.
• ›Adds endpoint_url support when using a boto3 session with DynamoDB, enabling custom or local DynamoDB endpoints.
• ›Moves OpenAI Assistants into LangChain core and adds callback support.

LangChain v0.0.335 adds FastEmbed embeddings, Neo4j chat history, a Docusaurus loader, and Cohere v3 embedding model support.

$ git clone --branch v0.0.335 https://github.com/langchain-ai/langchain.git
# already have the repo? check out this version:
$ git checkout v0.0.335

from langchain.embeddings import FastEmbedEmbeddings

embeddings = FastEmbedEmbeddings()
vectors = embeddings.embed_documents(["LangChain is a framework for LLM apps."])

• ›Adds FastEmbed embedding provider integration for fast, local embedding generation.
• ›Adds Neo4jChatMessageHistory for storing and retrieving chat message history in a Neo4j graph database.
• ›Adds DocusaurusLoader document loader to ingest content from Docusaurus-based documentation sites.
• ›Upgrades the Cohere embedding integration to use the v3 embedding model.
• ›Makes RunnableBinding easier to subclass with custom __init__ arguments.

LangChain v0.0.333 adds embeddings filter score state, Vertex AI snippet retrieval, and OpenAI tool improvements.

$ git clone --branch v0.0.333 https://github.com/langchain-ai/langchain.git
# already have the repo? check out this version:
$ git checkout v0.0.333

• ›Adds embeddings filter option to return similarity scores in retriever state, enabling downstream score-aware processing.
• ›Adds snippet retrieval support for non-advanced website data stores in Vertex AI Search.
• ›Adds a Tool Retrieval prompt template for dynamic tool selection workflows.
• ›Adds ability to convert Cohere chat messages to LangChain documents.

LangChain v0.0.332 adds Astra DB vector store, Cohere Embed v3, OpenAI Assistants, and new RAG templates.

$ git clone --branch v0.0.332 https://github.com/langchain-ai/langchain.git
# already have the repo? check out this version:
$ git checkout v0.0.332

• ›Adds Memorize tool, enabling agents to store information to long-term memory during a conversation.
• ›Adds support for Cohere Embed v3 embeddings.
• ›Adds 'Astra DB' vector store integration.
• ›Adds OpenAI Assistants support, including multiple actions per assistant.
• ›Records system_fingerprint field on ChatOpenAI responses.

LangChain v0.0.331rc3 adds Astra DB vector store, Memorize tool, OAI assistant multi-action support, Neo4j templates, and Azure OpenAI Embeddings.

$ git clone --branch v0.0.331rc3 https://github.com/langchain-ai/langchain.git
# already have the repo? check out this version:
$ git checkout v0.0.331rc3

• ›Adds Memorize tool, enabling agents to write information into long-term memory during a session.
• ›Adds Astra DB vector store integration for using DataStax Astra DB as a vector backend.
• ›Adds Azure OpenAI Embeddings integration.
• ›Adds OpenAI Assistant support for multiple actions in a single run.
• ›Adds a Neo4j conversation Cypher template for graph-based conversational retrieval.

LangChain v0.0.331rc2 adds OpenAI v1 embeddings support, a Vectara RAG template, and MongoDB ingest.

$ git clone --branch v0.0.331rc2 https://github.com/langchain-ai/langchain.git
# already have the repo? check out this version:
$ git checkout v0.0.331rc2

• ›Adds OpenAI v1 embeddings support.
• ›Adds a Vectara RAG template for retrieval-augmented generation pipelines.
• ›Adds MongoDB ingest support.

LangChain v0.0.331rc0 adds Cohere Embed v3 support, OpenAI system fingerprint recording, and per-conversation artifact callbacks.

$ git clone --branch v0.0.331rc0 https://github.com/langchain-ai/langchain.git
# already have the repo? check out this version:
$ git checkout v0.0.331rc0

• ›Adds support for Cohere Embed v3 embeddings.
• ›Records the OpenAI system fingerprint in ChatOpenAI responses.
• ›Adds on_artifacts callback parameter to pass artifact handlers for a specific conversation.

LangChain v0.0.331 adds MongoDB parent document retrieval support.

$ git clone --branch v0.0.331 https://github.com/langchain-ai/langchain.git
# already have the repo? check out this version:
$ git checkout v0.0.331

• ›Adds MongoDB parent document retrieval, enabling ParentDocumentRetriever backed by Mongo storage.

LangChain v0.0.330 adds pgvecto.rs and TileDB vector stores, Zep summary search, OpenCLIP multimodal embeddings, and new RAG templates.

$ git clone --branch v0.0.330 https://github.com/langchain-ai/langchain.git
# already have the repo? check out this version:
$ git checkout v0.0.330

• ›Enables the device_map parameter in the HuggingFace pipeline integration.
• ›Adds pgvecto.rs as a new VectorStore backend.
• ›Adds TileDB as a new VectorStore implementation.
• ›Adds Zep summary search capability with accompanying usage example.
• ›Adds native MMR (Maximal Marginal Relevance) support to the Zep VectorStore.

LangChain v0.0.329 adds Runnable.with_listeners(), bind_functions(), LM Format Enforcer integration, Quip loader, and a version CLI command.

$ git clone --branch v0.0.329 https://github.com/langchain-ai/langchain.git
# already have the repo? check out this version:
$ git checkout v0.0.329

from langchain.chat_models import ChatOpenAI

functions = [
    {
        'name': 'get_weather',
        'description': 'Get current weather for a city',
        'parameters': {
            'type': 'object',
            'properties': {'city': {'type': 'string'}},
            'required': ['city']
        }
    }
]

llm_with_fns = ChatOpenAI(model='gpt-4').bind_functions(functions)
llm_with_fns.invoke('What is the weather in Paris?')

• ›Adds Runnable.with_listeners() method to attach event listeners to any Runnable in a chain.
• ›Adds bind_functions() convenience method on Runnable for binding callable functions directly.
• ›Adds version subcommand to the langchain-cli for inspecting the installed CLI version.
• ›Adds LM Format Enforcer integration for structured/constrained LLM output.
• ›Adds Quip document loader for ingesting Quip content.

Letta 0.2.3 adds configurable presets, a WebSocket server interface, and version-tracked agent configs.

$ git clone --branch 0.2.3 https://github.com/letta-ai/letta.git
# already have the repo? check out this version:
$ git checkout 0.2.3

$ memgpt configure

• ›Adds memgpt_version field to stored configs so agents track which version they were saved with, improving cross-version compatibility.
• ›Adds load and load_and_attach functions to the MemGPT AutoGen agent integration.
• ›Introduces a WebSocket interface via server.py for real-time agent communication.
• ›Introduces configurable presets, letting developers customize the function set and system prompts MemGPT agents use.

Letta 0.2.0 adds pgvector archival memory, Ollama support, grammar-based sampling, file I/O, and new CLI commands.

$ git clone --branch 0.2.0 https://github.com/letta-ai/letta.git
# already have the repo? check out this version:
$ git checkout 0.2.0

$ memgpt version

$ memgpt run --persona sam --human user --model gpt-4-turbo

• ›Adds MEMGPT_CONFIG_PATH environment variable to override the default config location (~/.memgpt/config).
• ›Adds memgpt version command to print the installed package version.
• ›Adds /retry in-chat command to request another answer from the agent.
• ›Adds pgvector (PostgreSQL vector database) support for archival memory storage.
• ›Adds Ollama as a supported local LLM backend.

LlamaIndex v0.9.9 adds LlamaDataset abstractions, metadata filtering, and MMR mode for AstraDB vector store.

$ git clone --branch v0.9.9 https://github.com/run-llama/llama_index.git
# already have the repo? check out this version:
$ git checkout v0.9.9

• ›Adds metadata filtering and MMR (Maximal Marginal Relevance) mode support to AstraDBVectorStore.
• ›Introduces new abstractions for LlamaDataset in the evaluation module.
• ›Allows latest scikit-learn versions as a compatible dependency.

LlamaIndex v0.9.8 adds async metadata extraction, ObjectIndex persistence, and character-index tracking in nodes.

$ git clone --branch v0.9.8 https://github.com/run-llama/llama_index.git
# already have the repo? check out this version:
$ git checkout v0.9.8

• ›Adds persist and persist_from_dir methods to ObjectIndex for saving and restoring index state to disk.
• ›Adds async metadata extraction with pipeline support for non-blocking ingestion workflows.

Semantic Kernel Python adds Azure CosmosDB Mongo vCore memory store, pre/post RunAsync handlers, and OpenAI user-agent headers.

$ git clone --branch python-0.3.15.dev https://github.com/microsoft/semantic-kernel.git
# already have the repo? check out this version:
$ git checkout python-0.3.15.dev

• ›Adds Azure CosmosDB Mongo vCore as a vector memory datastore, expanding the set of supported backends for semantic memory.
• ›Syncs pre/post RunAsync event handlers from C# to Python, enabling hook-based pipeline instrumentation around kernel function execution.
• ›Adds a user-agent header to all OpenAI and OpenAPI HTTP requests, improving traceability of Semantic Kernel traffic at the API gateway level.

Jan v0.3.1 adds an experimental feature toggle, a new conversation button, and a revamped plugin architecture.

$ git clone --branch v0.3.1 https://github.com/janhq/jan.git
# already have the repo? check out this version:
$ git checkout v0.3.1

• ›Adds an Experimental Feature Toggle, making opt-in unstable features accessible without separate builds.
• ›Adds a 'New Conversation' button to the conversation sidebar for faster session creation.
• ›Revamps the plugin architecture, including refactoring the plugin manager and execution layer to TypeScript.
• ›Reads plugin manifests from a CDN, enabling dynamic plugin discovery without local manifest files.
• ›Adds appDataPath to the plugin SDK for plugins that need access to the application data directory.

KoboldCpp v1.49 adds Split Memory, trim_stop, and --preloadstory API features for richer generation control.

$ git clone --branch v1.49 https://github.com/LostRuins/koboldcpp.git
# already have the repo? check out this version:
$ git checkout v1.49

$ curl -X POST http://localhost:5001/v1/generate \
  -H 'Content-Type: application/json' \
  -d '{"prompt": "The adventurer enters the dungeon.", "memory": "You are a dungeon master. The setting is a dark fantasy world.", "max_length": 200}'

$ curl -X POST http://localhost:5001/v1/generate \
  -H 'Content-Type: application/json' \
  -d '{"prompt": "Once upon a time", "stop_sequence": ["###", "END"], "trim_stop": true, "max_length": 150}'

$ koboldcpp.exe --model my_model.gguf --preloadstory my_save.json

• ›Adds memory field to the /v1/generate API payload: forcefully prepends a string to any submitted prompt, and if the context limit is exceeded, overwrites from the beginning of the main prompt to make room — guaranteeing full memory insertion without needing exact token counts.
• ›Adds trim_stop boolean field to the generate API payload: when true, strips detected stop sequences from the output and truncates everything after them (note: incompatible with SSE streaming).
• ›Adds --preloadstory CLI flag to specify a JSON story savefile at server launch, hosting it at the /api/extra/preloadstory endpoint for frontends to consume over the API.
• ›Adds LLAMA_PORTABLE=1 makefile flag for building portable binaries targeting Colab or Docker environments.
• ›Expands Kobold Lite with World Info inject position support, Split Memory, preloaded stories, and optional image generation via DALL-E 3 (OpenAI API).

KoboldCpp v1.48.1 adds KV cache context shifting, Min-P sampler, remote tunneling, and GPU auto-configuration.

$ git clone --branch v1.48.1 https://github.com/LostRuins/koboldcpp.git
# already have the repo? check out this version:
$ git checkout v1.48.1

$ koboldcpp.exe --model my_model.gguf --contextsize 8192
# Context Shifting is on by default; to opt out:
koboldcpp.exe --model my_model.gguf --contextsize 8192 --noshift

$ koboldcpp.exe --model my_model.gguf --remotetunnel

• ›Adds --noshift flag to disable the new Context Shifting (EvenSmarterContext) feature, which uses KV cache shifting to remove old tokens and add new ones without reprocessing — enabled by default and overrides SmartContext when both are set.
• ›Adds --remotetunnel flag, which downloads Cloudflared and creates a TryCloudFlare tunnel so KoboldCpp is reachable over the internet even behind a firewall.
• ›Adds Min-P sampler, now available via the API and configurable in Kobold Lite under the Advanced settings tab.
• ›Introduces a new build target koboldcpp_clblast_noavx2 ('CLBlast NoAVX2 (Old CPU)') for Windows users without AVX2 intrinsics, enabling CLBlast GPU acceleration on older CPUs.
• ›Changes MMQ/Tensor Core behavior: MMQ is always enabled until batch > 32, CuBLAS only activates for larger batches when the MMQ flag is explicitly disabled, and MMQ dimensions are set to 'FAVOR BIG' — diverging from upstream llama.cpp's approach.

OpenAI API becomes the default, gains /v1/internal/stop-generation endpoint, and now supports trust_remote_code for embeddings.

$ git clone --branch snapshot-2023-11-12 https://github.com/oobabooga/textgen.git
# already have the repo? check out this version:
$ git checkout snapshot-2023-11-12

$ curl -X POST http://localhost:5000/v1/internal/stop-generation

• ›Adds POST /v1/internal/stop-generation endpoint to the OpenAI-compatible API, allowing programmatic cancellation of in-progress generation.
• ›Makes the OpenAI-compatible API the default API (previously non-default).
• ›Enables trust_remote_code support in the OpenAI API embedder, allowing embedding models that require remote code execution.
• ›Separates context and system message fields in instruction formats, enabling independent control of each in prompt templates.

Adds Min P sampler, temperature_last parameter, and use_flash_attention_2 flag to oobabooga text-generation-webui.

$ git clone --branch snapshot-2023-11-05 https://github.com/oobabooga/textgen.git
# already have the repo? check out this version:
$ git checkout snapshot-2023-11-05

• ›Adds temperature_last parameter to control the order in which temperature sampling is applied relative to other samplers.
• ›Adds use_flash_attention_2 parameter to the Transformers model loader to enable Flash Attention 2 support.
• ›Implements Min P as a new sampler option in HF loaders for controlling minimum probability thresholds during generation.
• ›Adds a flag to force loading models from safetensors format in the Transformers loader.

Ollama v0.1.12 adds Yi Chat 34B and improves multi-line prompt handling in the CLI.

$ git clone --branch v0.1.12 https://github.com/ollama/ollama.git
# already have the repo? check out this version:
$ git checkout v0.1.12

• ›Adds Yi Chat (the chat variant of the Yi 34B model) to the Ollama model library, available via ollama run yi.
• ›Supports multi-line prompts delimited by """ and improved paste functionality in ollama run.
• ›Adds Option (Alt) + Backspace word-deletion keybinding in the ollama run interactive prompt.

Ollama v0.1.11 adds Orca 2, DeepSeek Coder, and Alfred models plus GPU support for q5_0 and q5_1 quantizations.

$ git clone --branch v0.1.11 https://github.com/ollama/ollama.git
# already have the repo? check out this version:
$ git checkout v0.1.11

• ›Adds q5_0 and q5_1 quantized models to GPU execution, unlocking faster inference for those quantization levels.
• ›Adds Orca 2 model (orca2), a Llama 2 fine-tune optimized for reasoning tasks.
• ›Adds DeepSeek Coder model (deepseek-coder), a code-focused model available in 1.3B, 6.7B, and 33B parameter sizes.
• ›Adds Alfred model (alfred), a conversational model supporting both chat and instruct use cases.

Ollama v0.1.10 adds JSON mode for ollama run, stdin prompt piping, and remote model builds via OLLAMA_HOST.

$ git clone --branch v0.1.10 https://github.com/ollama/ollama.git
# already have the repo? check out this version:
$ git checkout v0.1.10

$ head -30 README.md | ollama run codellama 'how do I install Ollama on Linux?'

• ›Adds --format json flag to ollama run to enable JSON output mode from the command line.
• ›Adds /set format json in-session command to switch an active ollama run chat session to JSON mode.
• ›ollama create now respects OLLAMA_HOST to build models against a remote Ollama instance.
• ›Enables piping prompts into ollama run via standard input, allowing shell pipelines like head -30 README.md | ollama run codellama '<question>'.
• ›Adds three new models to the library: OpenChat (ollama run openchat), Neural-chat (ollama run neural-chat), and Goliath (ollama run goliath).

Ollama v0.1.9 adds JSON mode and raw mode to /api/generate, plus a new bilingual Yi model.

$ git clone --branch v0.1.9 https://github.com/ollama/ollama.git
# already have the repo? check out this version:
$ git checkout v0.1.9

$ curl http://localhost:11434/api/generate -d '{"model": "llama2", "prompt": "List the top 3 open ports on a typical web server.", "format": "json"}'

$ curl http://localhost:11434/api/generate -d '{"model": "llama2", "prompt": "[INST] Summarize this CVE. [/INST]", "raw": true}'

• ›Adds format parameter to POST /api/generate — set it to json to force models to always return valid JSON (JSON mode).
• ›Adds raw parameter to POST /api/generate — set {"raw": true} to bypass Ollama's prompt templating entirely (raw mode).
• ›Adds the Yi bilingual model (English and Chinese) to the Ollama library, available via ollama pull yi.

Ollama v0.1.8 adds five new models and dramatically faster push speeds up to 1 GB/s for large models.

$ git clone --branch v0.1.8 https://github.com/ollama/ollama.git
# already have the repo? check out this version:
$ git checkout v0.1.8

• ›Adds codebooga to the model library: a high-performing code instruct model created by merging two existing code models.
• ›Adds dolphin2.2-mistral to the model library: a Mistral-based instruct-tuned model fine-tuned for improved conversation and empathy.
• ›Adds mistrallite to the model library: a Mistral fine-tune with enhanced long-context processing capabilities.
• ›Ollama now honours large context sizes on models such as codellama and mistrallite.
• ›ollama push is now dramatically faster: 7B models push at up to ~100 MB/s and 70B+ models at up to 1 GB/s when network permits.

Phoenix v1.3.0 adds OpenAI rate limiting, async eval submission, configurable trace exporter endpoint, and richer evaluation UI across spans and documents.

$ git clone --branch v1.3.0 https://github.com/Arize-ai/phoenix.git
# already have the repo? check out this version:
$ git checkout v1.3.0

• ›Adds configurable endpoint for the trace exporter, letting users point spans at a custom collector address.
• ›Adds OpenAI rate limiting for eval calls, preventing API quota exhaustion during large evaluation runs.
• ›Implements asynchronous submission for OpenAI evals, enabling non-blocking evaluation workflows.
• ›Adds a reference link correctness evaluation prompt template for assessing retrieval quality.
• ›Shows span evaluations in the trace details slideout, surfacing eval scores and labels inline with trace data.

Arize Phoenix v1.2.0 adds LiteLLM model support for evals, partial result recovery, a Dockerfile, and SageMaker notebook support.

$ git clone --branch v1.2.0 https://github.com/Arize-ai/phoenix.git
# already have the repo? check out this version:
$ git checkout v1.2.0

• ›Adds LiteLLM model support for evals, enabling any LiteLLM-supported model to be used as the evaluation LLM.
• ›Evals now return partial results when the LLM function is interrupted mid-run, preserving completed evaluation outputs.
• ›Adds a Dockerfile for containerized deployment of Phoenix.
• ›Adds SageMaker notebook support, enabling Phoenix to run in Amazon SageMaker environments.

Phoenix v1.1.0 adds explanation output to evals and an output_parser argument to llm_generate.

$ git clone --branch v1.1.0 https://github.com/Arize-ai/phoenix.git
# already have the repo? check out this version:
$ git checkout v1.1.0

llm_generate(dataframe=df, template=my_template, model=model, output_parser=my_parser)

• ›Adds output_parser argument to llm_generate for custom parsing of LLM responses during eval generation.
• ›Evals now support explanations, surfacing model reasoning alongside evaluation labels.

Arize Phoenix v1.0.0 adds OpenAI 1.0 SDK support, with a breaking change for existing model integrations.

$ git clone --branch v1.0.0 https://github.com/Arize-ai/phoenix.git
# already have the repo? check out this version:
$ git checkout v1.0.0

• ›Adds support for the OpenAI 1.0 SDK in the models integration layer.

Arize Phoenix v0.1.0 adds long-context evaluators with map-reduce and refine patterns, plus span table column visibility controls.

$ git clone --branch v0.1.0 https://github.com/Arize-ai/phoenix.git
# already have the repo? check out this version:
$ git checkout v0.1.0

• ›Adds long-context evaluators supporting map-reduce and refine patterns for evaluating LLM outputs that exceed standard context windows.
• ›Adds column visibility controls to the span table in the traces UI, letting practitioners show or hide columns.

Langfuse v1.9.0 lets you invite new users to a project before they have an account, with optional email notifications via SMTP.

$ git clone --branch v1.9.0 https://github.com/langfuse/langfuse.git
# already have the repo? check out this version:
$ git checkout v1.9.0

$ [email protected]
SMTP_CONNECTION_URL=smtp://user:[email protected]:587

• ›Adds EMAIL_FROM_ADDRESS and SMTP_CONNECTION_URL environment variables to enable email notifications when inviting new (accountless) users to a project.
• ›Supports inviting users to a project who do not yet have a Langfuse account.

Langfuse v1.8.2 adds project name display to the UI breadcrumb.

$ git clone --branch v1.8.2 https://github.com/langfuse/langfuse.git
# already have the repo? check out this version:
$ git checkout v1.8.2

• ›Shows the project name in the UI breadcrumb for easier navigation across projects.

Langfuse v1.8.0 adds the ability to bookmark traces in the UI.

$ git clone --branch v1.8.0 https://github.com/langfuse/langfuse.git
# already have the repo? check out this version:
$ git checkout v1.8.0

• ›Adds bookmark support for traces in the UI, with bookmark state cached in the component for fast interaction.

Langfuse v1.7.0 adds a projects API endpoint for SDK-side API key validation.

$ git clone --branch v1.7.0 https://github.com/langfuse/langfuse.git
# already have the repo? check out this version:
$ git checkout v1.7.0

• ›Adds a projects API endpoint to validate API keys from within SDKs.

Langfuse v1.6.0 adds 'does not contain' string filters, SSO on sign-up, and a data region switch for Cloud.

$ git clone --branch v1.6.0 https://github.com/langfuse/langfuse.git
# already have the repo? check out this version:
$ git checkout v1.6.0

• ›Adds 'does not contain' filter option for string fields in the UI, enabling negative-match filtering across traces and other string-based views.
• ›Surfaces SSO options on the /auth/sign-up page so users can discover and use single sign-on at registration time.
• ›Adds a data region switch for Langfuse Cloud, letting users select their preferred hosting region.

Langfuse v1.5.0 adds an env var to enforce SSO-only login and disable username/password auth.

$ git clone --branch v1.5.0 https://github.com/langfuse/langfuse.git
# already have the repo? check out this version:
$ git checkout v1.5.0

$ AUTH_DISABLE_USERNAME_PASSWORD=true

• ›Adds AUTH_DISABLE_USERNAME_PASSWORD environment variable to enforce SSO-only authentication, preventing username/password login.

Langfuse v1.4.0 adds the ability to transfer project ownership to a new owner.

$ git clone --branch v1.4.0 https://github.com/langfuse/langfuse.git
# already have the repo? check out this version:
$ git checkout v1.4.0

• ›Adds 'Transfer project to new owner' capability, allowing project ownership to be reassigned within Langfuse.

Langfuse v1.3.0 adds single-trace deletion in the UI, a health check API endpoint, and project name display in the header.

$ git clone --branch v1.3.0 https://github.com/langfuse/langfuse.git
# already have the repo? check out this version:
$ git checkout v1.3.0

$ curl -f https://<your-langfuse-host>/api/public/health

• ›Adds GET /api/public/health endpoint for health-check monitoring of the Langfuse service.
• ›Enables deletion of individual traces directly from the UI.
• ›Displays the project name in the application header for clearer multi-project context.
• ›Removes the separate cron process used for telemetry in the Docker deployment, simplifying the container footprint.

Langfuse v1.2.0 adds more string filters to the UI.

$ git clone --branch v1.2.0 https://github.com/langfuse/langfuse.git
# already have the repo? check out this version:
$ git checkout v1.2.0

• ›Adds more string filter options to the UI for querying traces and observations.

Langfuse v1.1.0 adds upsert support for scores via the API.

$ git clone --branch v1.1.0 https://github.com/langfuse/langfuse.git
# already have the repo? check out this version:
$ git checkout v1.1.0

• ›Adds upsert support for scores via the API, allowing existing score records to be created or updated in a single call.

Chroma 0.4.17 adds OpenAI v1.x support and a new system-catalog-provider for simpler deployments.

$ git clone --branch 0.4.17 https://github.com/chroma-core/chroma.git
# already have the repo? check out this version:
$ git checkout 0.4.17

• ›Supports OpenAI Python package v1.x.x in utils.OpenAIEmbeddingFunction, alongside a new deployment_id parameter for the v0.x.x API.
• ›Adds system-catalog-provider configuration option to simplify distributed/multi-node deployment setup.

Chroma 0.4.16 adds authorization (authz) support and multimodal embedding functions.

$ git clone --branch 0.4.16 https://github.com/chroma-core/chroma.git
# already have the repo? check out this version:
$ git checkout 0.4.16

• ›Adds authorization (authz) framework with resource attribute extraction for tenant, database, and list_collections operations, mapping identity attributes to AuthzUser.
• ›Adds multimodal embedding functions, enabling embeddings to be generated from multiple modalities (e.g., image and text) within Chroma.
• ›Improves HTTPClient connection error messages to surface clearer diagnostics when the server is unreachable.

LanceDB v0.3.8 adds SaaS create_index API and exposes index cache size in Python.

$ git clone --branch v0.3.8 https://github.com/lancedb/lancedb.git
# already have the repo? check out this version:
$ git checkout v0.3.8

• ›Exposes index cache size configuration in the Python client (feat(python): expose index cache size).
• ›Adds a create_index API for SaaS (cloud-hosted) LanceDB deployments.

LanceDB v0.3.7 adds exponential backoff for rate-limited embeddings, multi-task Instructor model with quantization, and RemoteTable.version in Python.

$ git clone --branch v0.3.7 https://github.com/lancedb/lancedb.git
# already have the repo? check out this version:
$ git checkout v0.3.7

• ›Adds RemoteTable.version property in the Python SDK to retrieve the current version of a remote table.
• ›Adds exponential backoff retry support for embedding functions that hit rate limits.
• ›Adds multi-task Instructor model support with quantization, plus a weak_lru cache for embedding function models to reduce redundant model loads.

LanceDB v0.3.6 adds prefilter support for ANN index queries.

$ git clone --branch v0.3.6 https://github.com/lancedb/lancedb.git
# already have the repo? check out this version:
$ git checkout v0.3.6

• ›Adds prefilter flag to allow prefiltering with an index during approximate nearest neighbor queries, enabling filtered vector search without a post-filter pass.

LanceDB v0.3.3 adds optimize/remap index APIs, dataset stats APIs, and prefilter support for indexed queries.

$ git clone --branch python-v0.3.3 https://github.com/lancedb/lancedb.git
# already have the repo? check out this version:
$ git checkout python-v0.3.3

• ›Adds optimize_index API to allow index optimization on existing tables.
• ›Adds remap_index API to support index remapping operations.
• ›Adds data/dataset stats APIs for retrieving dataset statistics (exposed in both Python and Node SDKs).
• ›Adds prefilter flag to allow prefiltering with an index during queries.

Milvus 2.2.15 adds bulkinsert support for partitionkey and pure-list JSON format, and removes MySQL metastore.

$ git clone --branch v2.2.15 https://github.com/milvus-io/milvus.git
# already have the repo? check out this version:
$ git checkout v2.2.15

• ›Enables bulkinsert of binlog data with partitionkey, allowing bulk imports into partition-key-enabled collections.
• ›Adds support for bulkinsert with pure list JSON format, expanding the accepted input formats for bulk data ingestion.

Milvus v2.3.3 adds pure list JSON support in bulk insert and improves rolling upgrade reliability.

$ git clone --branch v2.3.3 https://github.com/milvus-io/milvus.git
# already have the repo? check out this version:
$ git checkout v2.3.3

• ›Supports pure list JSON format in bulk insert operations.

Weaviate v1.22.3 adds the text2vec-jinaai module, Cohere v3 model support, and OpenAI GPT-4 128k/GPT-3.5 preview models.

$ git clone --branch v1.22.3 https://github.com/weaviate/weaviate.git
# already have the repo? check out this version:
$ git checkout v1.22.3

• ›Adds support for overriding Cohere's base URL via the X-Cohere-BaseURL HTTP header, enabling routing to custom or proxy endpoints.
• ›Adds support for passing OpenAI base URL via HTTP header, enabling routing to custom or proxy OpenAI-compatible endpoints.
• ›Adds the text2vec-jinaai module, enabling JinaAI embeddings as a vectorization source in Weaviate.
• ›Adds support for Cohere v3 models in the Cohere integration.
• ›Adds support for OpenAI GPT-4 128k and GPT-3.5 preview models in the OpenAI integration.