Liking cljdoc? Tell your friends :D
Clojure only.

bosquet.agent.react

focus-on-observationclj

(focus-on-observation {:keys [lookup-db lookup-index]})

Get the sentence a the position lookup-index from the observation.

Get the sentence a the position `lookup-index` from the observation.
raw docstring

generate-thoughtsclj

(generate-thoughts ctx prompt-palette prompt-key)

Generate ReAct thoughts. ctx has needed data points prompt-palette has all the prompts needed for ReAct prompt-key is the key to the prompt to be used to start the generation.

Generate ReAct thoughts.
`ctx` has needed data points
`prompt-palette` has all the prompts needed for ReAct
`prompt-key` is the key to the prompt to be used to start the generation.
raw docstring

solve-taskclj

(solve-task tool
            prompt-palette
            {:keys [task max-steps] :or {max-steps 5} :as initial-ctx})

Solve a task using ReAct

First agent parameter specifies which tool will be used to solve the task. Second context parameter gives initialization data to start working

  • task is a quesiton ar task formulation for the agent
  • max-steps specifies how many thinking steps agent is allowed to do it either reaches that number of steps or 'Finish' action, and then terminates.

:react/task contains a question or a claim to be solved

Solve a task using [ReAct](https://react-lm.github.io)

First `agent` parameter specifies which tool will be used to solve the task.
Second context parameter gives initialization data to start working
- `task` is a quesiton ar task formulation for the agent
- `max-steps` specifies how many thinking steps agent is allowed to do
it either reaches that number of steps or 'Finish' action, and then terminates.

:react/task contains a question or a claim to be solved
raw docstring

cljdoc is a website building & hosting documentation for Clojure/Script libraries

Keyboard shortcutsReport a problemcljdoc on GitHub
× close