Packages

beam_weaver

0.1.6
0.1.13 0.1.12 0.1.11 0.1.10 0.1.9 0.1.8 0.1.7 0.1.6 0.1.5 0.1.4 0.1.3 0.1.2 0.1.1 0.1.0

Elixir-native LangChain, LangGraph, and DeepAgents for traceable LLM apps: OTP workflows, tools, memory, human-in-the-loop, streaming, custom clients/adapters, minimal deps, and WeaveScope tracing.

Current section

Files

Jump to
Readme
14 Versions
7 Dependencies
Dependants
Files
Activity
beam_weaver examples prompt_caching.exs
Raw

examples/prompt_caching.exs 

Code.require_file("support.exs", __DIR__)



alias BeamWeaver.Core.ChatModel

alias BeamWeaver.Core.Message

alias BeamWeaver.Examples.Support



model = Support.model()



system_prompt = """

You are a support analysis agent.



Use the support policy below when answering. The policy is intentionally long

and stable so repeated calls can reuse provider prompt caches.



#{String.duplicate("- Verify account ownership, preserve audit trails, cite the support rule id, and never invent data.\n", 240)}

"""



messages =

  Support.prompt_cache_messages(model, system_prompt, [

    Message.user("Ticket SUP-42 asks whether deleted exports can be restored.")

  ])



cache_opts = Support.prompt_cache_opts(model, "support-agent", system_prompt)



{:ok, first} = ChatModel.invoke(model, messages, cache_opts)

{:ok, second} = ChatModel.invoke(model, messages, cache_opts)



IO.puts(Message.text(second))

IO.puts("")

IO.puts("First cached input tokens: #{Support.cache_read_tokens(first)}")

IO.puts("Second cached input tokens: #{Support.cache_read_tokens(second)}")