add harness, framework, runtime comparison #2319
Conversation
src/oss/concepts/products.mdx
Outdated
| | Skills | [Multi-agent skills](/oss/langchain/multi-agent/skills) | - | [Skills](/oss/deepagents/skills) | | ||
| | Subagents | [Multi-agents subagents](/oss/langchain/multi-agent/subagents) | [Subgraphs](/oss/langgraph/use-subgraphs) | [Subagents](/oss/deepagents/subagents) | | ||
| | Human-in-the-loop | [Human-in-the-loop middleware](/oss/langchain/human-in-the-loop) | [Interrupts](/oss/langgraph/interrupts) | [`interrupt_on` parameter](/oss/deepagents/harness#human-in-the-loop) | | ||
| | Streaming | [Agent Streaming](/oss/langchain/streaming/overview) | [Streaming](/oss/langgraph/streaming) | [Subagents streaming](/oss/deepagents/subagents#streaming) | |
There was a problem hiding this comment.
for deepagents, should be more than just subagents stremaing
src/oss/concepts/products.mdx
Outdated
| | Short-term memory | [Short-term memory](/oss/langchain/short-term-memory) | [Short-term memory](/oss/langgraph/add-memory#add-short-term-memory) | [`StateBackend`](/oss/deepagents/backends#statebackend-ephemeral) | | ||
| | Long-term memory | [Long-term memory](/oss/langchain/long-term-memory) | [Long-term memory](/oss/langgraph/add-memory#add-long-term-memory) | [Long-term memory](/oss/deepagents/long-term-memory) | | ||
| | Skills | [Multi-agent skills](/oss/langchain/multi-agent/skills) | - | [Skills](/oss/deepagents/skills) | | ||
| | Subagents | [Multi-agents subagents](/oss/langchain/multi-agent/subagents) | [Subgraphs](/oss/langgraph/use-subgraphs) | [Subagents](/oss/deepagents/subagents) | |
There was a problem hiding this comment.
multi-agents -> multi-agent ?
should be consistent with above
| Supported tools may include: | ||
|
|
||
| - **Durable execution**: Agents persist through failures and can run for extended periods, resuming from where they left off | ||
| - **Streaming**: Support for streaming workflows and responses |
There was a problem hiding this comment.
Since LangChain packages are built on top of LangGraph this is not specific to LangGraph only but also available in same capacity in LangChain and deepagent
There was a problem hiding this comment.
I think you're not wrong that LangChain has HIL, however I think the point I got from Harrison's blog post is that this capability is provided by the (underlying) runtime? Maybe I should not say "supported tools may include" but rather "Provided capabilities may include" - do you think that makes sense?
|
|
||
| - **Durable execution**: Agents persist through failures and can run for extended periods, resuming from where they left off | ||
| - **Streaming**: Support for streaming workflows and responses | ||
| - **Human-in-the-loop**: Incorporate human oversight by inspecting and modifying agent state |
| Use LangGraph when: | ||
|
|
||
| - You need durable execution for long-running, stateful agents | ||
| - You need fine-grained, low-level control over agent orchestration |
There was a problem hiding this comment.
i would put this above the previous bullet
Co-authored-by: Mason Daugherty <github@mdrxy.com>
| - These repeat every turn without caching | ||
| - Caching provides ~10x speedup and cost reduction | ||
|
|
||
| ## Streaming |
There was a problem hiding this comment.
@sydney-runkle do you mind reviewing this part?
| - **File system**: Read and write access to files on different pluggable storage backends | ||
| - **Token management**: Conversation history summarization and large tool result eviction | ||
|
|
||
| The [Deep Agents SDK](/oss/deepagents/overview) builds on top of LangChain and adds planning capabilities, file systems for context management, the ability to spawn subagents, and more. |
| ## Agent harnesses (like the Deep Agents SDK) | ||
|
|
||
| Agent harnesses are agent frameworks with built-in tools and capabilities that make building sophisticated, long-running agents easier. | ||
| Supported tools may include: |
There was a problem hiding this comment.
can we mention that they're designed to be opinionated and "batteries included" so that users can get started with agents for complex tasks easily out of the box
|
|
||
| - You are building agents that run over long time periods | ||
| - You are building agents that need to handle complex, multi-step tasks | ||
| - You want to use predefined tools, prompts and subagents for |
There was a problem hiding this comment.
maybe also good to mention built-in primitives that are important like filesystems and bash execution and automated context engineering
5 participants
No description provided.