Comparison with Langchain?

[ibehnam]

I skimmed the documentation and notebooks, and I found some similarities between guidance and langchain. Can you please explain the reasons one should use guidance over langchain? More importantly, some of the features (e.g., token healing) seem not to be available for GPT-4 (or closed-source models in general). Is that true?

[NotBrianZach]

what are the similarites with langchain? Semantic Kernel seems more like their direct langchain competitor

[2214962083]

i think this repo more like a new language ---- prompt language, but seems not complete

1. prompt langauage is not only txt file
2. it should has highlight syntax
3. it should be control the ai think params and reply result
4. provide agent by plugin and can control the agent in prompt lang with for/if

[QuangBK]

In my opinion, Guidance is clear and more straightforward because of prompt language. You can control the output of LLMs, force them to follow your instructions. That's a big different.
I have some troubles to build an agent with Langchain, since LLMs sometimes distract and doesn't follow the instruction well.
You may check this explanation.
Feel free to compare Guidance and Langchain with two ReAct agents: Guidance Agent and Langchain Agent

[PenutChen]

In my view, Guidance and LangChain exhibit significant differences despite both interacting with large language models. Guidance is primarily geared towards streamlining the process of guiding LLMs during inference, simplifying interaction and control. On the other hand, LangChain prioritizes the construction of applications that leverage the composability of LLMs, focusing on building practical, usable solutions.

[meirm]

I liked guidance syntax using the handles. In Langchain you need to write code in the specific language that you are using (js, python, ...). It could be cool guidance could be called by langchain as an LLM or use guidance agent using langchain tools.
Following the POSIX spirit I celebrate the differences and encourage that each tool focus on one thing only enabling chaining and interaction. A clear interface between langchain and guidance would be beneficial for everyone.

[marcotcr]

We talked about writing a detailed comparison here multiple times, but gave up on it. On the whole, we think guidance is lower level and less opinionated (having said this, we're not experts on LangChain!). We obviously don't try to do a lot of what LangChain does (e.g. retrieval), and as mentioned above there is no reason why someone would have to pick one or the other (since we do different things).

[atodniAr]

How about guidance and Semantic Kernel? We can expect integration between these two do we?

[aslisabanci]

There's microsoft/semantic-kernel#1550 to integrate guidance into semantic-kernel but it seems dormant.

[hongsw]

If only 'dotnet-build-and-test' passes the test, it seems to be deployed september 1st.

[kyeshmz]

@marcotcr is there a reason why it was given up? Even a small md in the README would give out helpful pointers other than the differences in corporate structure (MSFT vs MSFT Research)