docs: concepts for cloud and doc-reorg #2196

hwchase17 · 2024-10-26T23:02:56Z

Add conceptual information for langgraph cloud
Re-organize content

docs/docs/concepts/deployment.md

eyurtsev · 2024-10-27T00:52:22Z

docs/docs/concepts/deployment.md

+    return graph.invoke({...})
+```
+
+For simple applications, this is sufficient and is not that different from deploying Python applications in general, and so we will not go into this in too much detail here.


Maybe add 1 sentence noting any requirements from checkpointer or base store implementations?

Also worth clarifying that in memory saver should not be used?

not sure what to add?

i was thinking just clarifying (stating the obvious) that if you're deploying and need to get checkpointing, you'll probably want to use a postgres checkpointer and to actually get it working you'll need to get a postgres instance.

docs/docs/concepts/deployment.md

docs/docs/concepts/langgraph_platform.md

docs/docs/concepts/assistants.md

vbarda · 2024-10-28T14:51:59Z

docs/docs/concepts/assistants.md

+
+Once you've created an assistant, you can save and version it to track changes to the configuration over time. You can think about this at three levels:
+
+1) The graph lays out the general agent application logic 


i wonder if this can be brought up top, otherwise a user might get a bit lost since we're using this terminology (graph/agent/assistant) a bit interchangeably on this page

docs/docs/concepts/assistants.md

vbarda · 2024-10-28T14:55:33Z

docs/docs/concepts/double_texting.md

+This option rolls back all work done up until that point. 
+It then sends the user input in, basically as if it just followed the original run input. 
+
+This may create some weird states - for example, you may have two `User` messages in a row, with no `Asssitant` message in between them. 


i think we need to clarify here that we're talking about roles/ message types, otherwise some people might get confused on Assistant term collision

docs/docs/concepts/deployment.md

vbarda · 2024-10-28T15:12:36Z

docs/docs/concepts/deployment.md

+
+## Memory
+
+In addition to thread-level persistence (covered above by [checkpointers](#checkpointers)) you may need to deploy some storage to support cross-thread memory.


we should probably link to memory concepts here

docs/docs/concepts/langgraph_platform.md

eyurtsev · 2024-10-28T15:15:59Z

docs/docs/concepts/assistants.md

+2) The agent configuration options represent parameters that can be changed 
+3) Assistant versions save and track specific settings of the agent configuration options 
+
+For examples: let's imagine you have a general writing agent. You have created a general graph architecture that works well for writing. However - there are different types of writing, for examples blogs vs tweets. In order to get the best performance on each use case, you need to make some minor changes to the models and prompts used. In this setup, you could create an assistant for each use case - one for blog writing and one for tweeting. These would share the same graph structure, but they may use different models and different prompts. Read [this how-to](../cloud/how-tos/assistant_versioning.md) to learn how you can use assistant versioning through both the [Studio](../cloud/how-tos/index.md/#langgraph-studio) and the SDK.


OK I didn't see the new figures on mobile. The new figures make the version concept clear. The previous figures (the one currently up for versions) conflate versioning and configuration

eyurtsev · 2024-10-28T15:16:40Z

docs/docs/concepts/deployment.md

+# Deployment
+
+Once you build your LangGraph agent, you then have to think about deploying it.
+How you deploy your applications depends a bit on what it is and how you intend to use it.


nit: plural singular agreement

docs/docs/concepts/langgraph_platform.md

eyurtsev · 2024-10-28T15:22:07Z

docs/docs/concepts/langgraph_platform.md

+We've exposed endpoints for running your graph in the background.
+This includes endpoints for polling for the status of a run.
+
+**Support for long runs**


Maybe this should be consolidated with background run?

Another way this is called sometimes is "non-blocking" endpoints. These are usually created with the assumption that the underlying workload is heavy.

I think it goes further than that, eg. we've made sure endpoints don't time out when an agent takes more than 60s to produce output etc

say more? we're sending data in flight that reports status or some kind of heart beat to make sure connection doesn't get timed out?

eyurtsev · 2024-10-28T15:23:53Z

docs/docs/concepts/langgraph_platform.md

+
+There are several deployment options for LangGraph Platform.
+
+|                                | Free | Self Hosted | Bring Your Own Cloud | Cloud |


The free column is confusing.

Feels like we're mixing tiers vs. deployment options

eyurtsev · 2024-10-28T15:41:59Z

docs/docs/concepts/langgraph_platform.md

+
+### Cron Jobs
+
+It's often useful to run assistants on some schedule. LangGraph Cloud supports cron jobs, which run on a user defined schedule. The user specifies a schedule, an assistant, and some input. After than, on the specified schedule LangGraph cloud will:


needs 1 sentence example as motivation

eyurtsev · 2024-10-28T15:42:43Z

docs/docs/concepts/langgraph_platform.md

+
+### Cron Jobs
+
+It's often useful to run assistants on some schedule. LangGraph Cloud supports cron jobs, which run on a user defined schedule. The user specifies a schedule, an assistant, and some input. After than, on the specified schedule LangGraph cloud will:


"After that, on the specified schedule LangGraph cloud will:" fragment

eyurtsev · 2024-10-28T15:43:01Z

docs/docs/concepts/langgraph_studio.md

@@ -0,0 +1,33 @@
+# LangGraph Studio
+
+LangGraph Studio offers a new way to develop LLM applications by providing a specialized agent IDE that enables visualization, interaction, and debugging of complex agentic applications


LangGraph Studio is a specialized agent IDE that enables visualization, interaction, and debugging of complex agentic applications.

eyurtsev · 2024-10-28T15:45:13Z

docs/docs/concepts/langgraph_studio.md

@@ -0,0 +1,33 @@
+# LangGraph Studio


I don't know that the overview is motivating enough here.

How exactly does it improve collaboration with teammates?
Why is an IDE useful?

eyurtsev · 2024-10-28T15:47:36Z

docs/docs/concepts/streaming.md

+## LangGraph Platform
+
+Streaming is critical for making LLM applications feel responsive to end users. When creating a streaming run, the streaming mode determines what data is streamed back to the API client. LangGraph Platform supports five streaming modes.


This content feels duplicated since many of the modes should've already been documented?

Ideally we only document the differential

nfcampos · 2024-10-28T16:11:56Z

docs/docs/concepts/langgraph_platform.md

+**Background runs**
+
+We've exposed endpoints for running your graph in the background.
+This includes endpoints for polling for the status of a run.


As well as joining the real-time output stream of a background run.

nfcampos · 2024-10-28T16:12:54Z

docs/docs/concepts/langgraph_platform.md

+
+**Streaming endpoints**
+
+Endpoints that expose [multiple different streaming modes](streaming.md). We've made these work even for long running agents that may go minutes between steam event.


We also have support for re-joining the stream for an existing run if the connection drops, without interrupting the run

nfcampos · 2024-10-28T16:13:51Z

docs/docs/concepts/langgraph_platform.md

+| Who manages the infrastructure | You  | You         | Us                   | Us    |
+| With whom does the data reside | You  | You         | You                  | Us    |
+| LangGraph Studio Included?     | ❌    | ✅           | ✅                    | ✅     |
+| Assistants Included?           | ❌    | ✅           | ✅                    | ✅     |


Can we add crons/scheduled runs

docs/docs/concepts/langgraph_platform.md

docs/docs/concepts/deployment.md

eyurtsev · 2024-10-28T16:20:03Z

docs/docs/concepts/langgraph_platform.md

+We've exposed endpoints for running your graph in the background.
+This includes endpoints for polling for the status of a run.
+
+**Support for long runs**


say more? we're sending data in flight that reports status or some kind of heart beat to make sure connection doesn't get timed out?

docs/docs/concepts/deployment.md

Co-authored-by: Vadym Barda <[email protected]>

Co-authored-by: Nuno Campos <[email protected]>

docs: document template applications

Update platform / langgraph server documentation

Fix some formatting issues with CLI reference, add link to cli reference

update some links

Same content is already explained elsewhere

- fix broken links

Should resolve all broken links

concepts for cloud

63d3e33

eyurtsev reviewed Oct 27, 2024

View reviewed changes

vbarda reviewed Oct 28, 2024

View reviewed changes

docs/docs/concepts/assistants.md Outdated Show resolved Hide resolved

cr

18a9455

vbarda reviewed Oct 28, 2024

View reviewed changes

vbarda added 4 commits October 28, 2024 11:19

update

9e57f23

update

d05a884

update

197c4c7

update

e3d1c32

eyurtsev reviewed Oct 28, 2024

View reviewed changes

nfcampos reviewed Oct 28, 2024

View reviewed changes

docs/docs/concepts/langgraph_platform.md Outdated Show resolved Hide resolved

eyurtsev approved these changes Oct 28, 2024

View reviewed changes

eyurtsev reviewed Oct 28, 2024

View reviewed changes

docs/docs/concepts/deployment.md Outdated Show resolved Hide resolved

eyurtsev and others added 3 commits October 28, 2024 12:25

Apply suggestions from code review

d178b85

Co-authored-by: Vadym Barda <[email protected]>

Apply suggestions from code review

2a53197

Apply suggestions from code review

f361209

Co-authored-by: Nuno Campos <[email protected]>

eyurtsev changed the title ~~concepts for cloud~~ docs: concepts for cloud and doc-reorg Oct 28, 2024

eyurtsev self-requested a review October 28, 2024 16:27

vbarda and others added 9 commits October 29, 2024 11:40

format API ref

85b8bee

docs: add a how-to guide for using remote graph (#2218)

eb8e504

docs: update concepts and how-to guide indexes (#2203)

9cc9395

cr

8eafb10

nit

c734f44

add guide

c553a40

docs: update platform api ref (#2205)

5cd5aa8

cr

72bf29f

Merge pull request #2234 from langchain-ai/cc/template_apps

d61addc

docs: document template applications

ccurme and others added 26 commits October 30, 2024 15:28

docs: add webhooks to concepts (#2240)

62657af

docs: update cloud tutorial (#2233)

8c853f0

fix link

971bdcd

docs: update platform / langgraph server (#2231)

1d441c3

Update platform / langgraph server documentation

docs: update streaming concepts (#2239)

f4eeadb

format

da5369f

update links (#2241)

14dba26

Update CLI docs (#2238)

180a861

fix cli formatting issues (#2243)

02572f3

Fix some formatting issues with CLI reference, add link to cli reference

format

4d8467f

add sdk concept doc, more links (#2244)

0b6ef0f

merge references (#2245)

ffc8fc9

update some language (#2248)

2697898

docs: update cloud streaming guides (#2249)

a83f880

update API ref

478a86b

docs: update remote graph how-to (#2254)

5d3d13f

docs: update deployment plans (#2255)

6000900

typo

efce940

fix some links (#2257)

5037cf7

update some links

docs: fix docker-compose in self-hosted (#2258)

c7303da

remove deployment concept (#2259)

e22d387

Same content is already explained elsewhere

cr

2dccfba

cr

540a30b

docs: fix more broken links (#2260)

eb2cbf8

- fix broken links

docs: boom (#2261)

04d6ad9

Should resolve all broken links

Merge branch 'main' into harrison/concepts-cloud

c391cb0

eyurtsev merged commit d0e59f4 into main Oct 31, 2024
57 checks passed

eyurtsev deleted the harrison/concepts-cloud branch October 31, 2024 03:34

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

docs: concepts for cloud and doc-reorg #2196

docs: concepts for cloud and doc-reorg #2196

hwchase17 commented Oct 26, 2024 •

edited by eyurtsev

Loading

eyurtsev Oct 27, 2024

hwchase17 Oct 28, 2024

eyurtsev Oct 28, 2024

vbarda Oct 28, 2024

vbarda Oct 28, 2024

vbarda Oct 28, 2024

eyurtsev Oct 28, 2024

eyurtsev Oct 28, 2024

eyurtsev Oct 28, 2024

nfcampos Oct 28, 2024

eyurtsev Oct 28, 2024

eyurtsev Oct 28, 2024

eyurtsev Oct 28, 2024

eyurtsev Oct 28, 2024

eyurtsev Oct 28, 2024

eyurtsev Oct 28, 2024

eyurtsev Oct 28, 2024

nfcampos Oct 28, 2024

nfcampos Oct 28, 2024

nfcampos Oct 28, 2024

eyurtsev Oct 28, 2024


		Once you've created an assistant, you can save and version it to track changes to the configuration over time. You can think about this at three levels:

		1) The graph lays out the general agent application logic


		## Memory

		In addition to thread-level persistence (covered above by [checkpointers](#checkpointers)) you may need to deploy some storage to support cross-thread memory.


		There are several deployment options for LangGraph Platform.

		\| \| Free \| Self Hosted \| Bring Your Own Cloud \| Cloud \|


		### Cron Jobs

		It's often useful to run assistants on some schedule. LangGraph Cloud supports cron jobs, which run on a user defined schedule. The user specifies a schedule, an assistant, and some input. After than, on the specified schedule LangGraph cloud will:

		@@ -0,0 +1,33 @@
		# LangGraph Studio

		LangGraph Studio offers a new way to develop LLM applications by providing a specialized agent IDE that enables visualization, interaction, and debugging of complex agentic applications

		## LangGraph Platform

		Streaming is critical for making LLM applications feel responsive to end users. When creating a streaming run, the streaming mode determines what data is streamed back to the API client. LangGraph Platform supports five streaming modes.


		Streaming endpoints

		Endpoints that expose [multiple different streaming modes](streaming.md). We've made these work even for long running agents that may go minutes between steam event.

docs: concepts for cloud and doc-reorg #2196

docs: concepts for cloud and doc-reorg #2196

Conversation

hwchase17 commented Oct 26, 2024 • edited by eyurtsev Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

hwchase17 commented Oct 26, 2024 •

edited by eyurtsev

Loading