Using RunnablePassthrough/Parallel to Extract and Inject Values in a Chain

[ericwood73]

Checked other resources

• I added a very descriptive title to this question.
• I searched the LangChain documentation with the integrated search.
• I used the GitHub search to find a similar question and didn't find it.

Commit to Help

• I commit to help with one of those options 👆

Example Code

chain = RunnableParallel({
    "state": RunnablePassthrough(),
    "output": RunnableParallel({
        "state": RunnablePassthrough(),
        "input": before_llm
    }) | recovery_inject | RunnableParallel({
        "llm_input": RunnablePassthrough(),
        "output": llm | RunnableParallel({
            "llm_output": RunnablePassthrough(),
            "output": after_llm
        })
    })
})

Description

I'm working on a recovery wrapper using langgraph/langchain that can take a sequence of steps (each a chain) run them and if there is an error or a check fails, route the problem to an llm which can identify the step that introduced the problem, recommend corrective action, and route the error back to the step to correct. To do this I need a chain that can do several things at once:

1. Captures the input state to the chain and returns it as part of the output
2. Captures the input to the llm (after prompt) and the output from the llm (before output parsing) and makes it part of the output.
3. Modifies the input prompt to the LLM by injecting a recovery prompt that wraps the original prompt with corrective action to recover from an error. This will be a function that takes the state input to the chain and the current prompt and wraps the prompt if necessary.

This is the structure I have come up with for the chain. Do you see any way to simplify other than using intermediate variables to represent some of the structure? Am I using RunnableParallel and RunnablePassthrough correctly. I appreciate the help. I plan to open source this to the benefit of the langgraph/chain community once I have it stable.

System Info

System Information

OS: Darwin
OS Version: Darwin Kernel Version 22.5.0: Thu Jun 8 22:22:22 PDT 2023; root:xnu-8796.121.3~7/RELEASE_X86_64
Python Version: 3.11.5 (main, Aug 24 2023, 15:18:16) [Clang 14.0.3 (clang-1403.0.22.14.1)]

Package Information

langchain_core: 0.2.38
langchain: 0.2.16
langchain_community: 0.2.5
langsmith: 0.1.116
langchain_anthropic: 0.1.23
langchain_chroma: 0.1.3
langchain_cohere: 0.1.5
langchain_openai: 0.1.23
langchain_text_splitters: 0.2.4
langgraph: 0.2.19

Optional packages not installed

langserve

Other Dependencies

aiohttp: 3.10.5
anthropic: 0.34.2
async-timeout: Installed. No version info available.
chromadb: 0.5.3
cohere: 5.5.3
dataclasses-json: 0.6.6
defusedxml: 0.7.1
fastapi: 0.112.4
httpx: 0.27.2
jsonpatch: 1.33
langgraph-checkpoint: 1.0.9
numpy: 1.26.4
openai: 1.44.0
orjson: 3.10.7
packaging: 24.1
pydantic: 2.9.0
PyYAML: 6.0.2
requests: 2.32.3
SQLAlchemy: 2.0.34
tenacity: 8.5.0
tiktoken: 0.7.0
typing-extensions: 4.12.2