Seeking private testers for fluent - a new cli application to interact with FlowiseAI workflows and the linux pipeline.

[njfio]

Hi FlowiseAI community,

I'm developing a new open source command line tool called fluent (Flowise Language Utility for Enhanced Networked Transactions) private repo currently that allows you to easily interact with FlowiseAI workflows from the terminal. Fluent aims to provide a simple, streamlined interface for running workflows, handling input/output, and customizing behavior via an extensible configuration system.

Some key features of Fluent:

• Easily run any FlowiseAI workflow from the command line by name
• Combines question input, context, and outlines into a configurable prompt format
• Supports reading input context from stdin pipes for chaining with other CLI tools and more flowise workflows
• Outputs workflow results in plain text, Markdown, or a "structured" format extracting code blocks and special sections
• Allows overriding the system prompt on a per-workflow basis
• Can downloads and saves any images generated by the workflow
• Extensible configuration system for defining workflows and options
• Generates shell auto-completion scripts for easy integration and Flowise flow names.

The core logic is implemented but I'm looking for some private alpha testers to try it out, provide feedback, and help shake out any bugs before a wider release. You'd get early access to the tool to use in your own projects.

If you're interested in testing Fluent, please reply here or DM me. I'm looking for folks comfortable using command line tools who can provide detailed feedback. Familiarity with Go development is a plus for code review feedback but not required.

Thanks! I'm excited to share Fluent with the community and make FlowiseAI workflows more accessible and useful for everyone.

Nick

A few examples in use

Note: The preceding messaage was generated through a fluent workflow

cat src/*.rs | fluent AnthropicOpusXMLFunctionAgent "based on the information provided in the context, please write a compelling post about fluent, I want to post to the flowiseAI github discussion group and ask for anybody interested in private testing."

1. cat src/*.rs reads the contents of all the Rust source files in the src directory and pipes the output to the next command
2. | pipes the output of cat to the fluent command
3. fluent AnthropicOpusXMLFunctionAgent invokes the fluent command, passing it the AnthropicOpusXMLFunctionAgent argument and using the contents of 'stdin' in the 'context' of the question.
4. It asks the agent to write a compelling post about fluent based on the context provided (which is the contents of the Rust source files).

An Example Interview Script

Attached below is an example interview with a GPT4 agent interviewed Another GPT4 Turbo agent, Gemini, Perplexity, Mixtral, GrokOpenLLama, and AnthropicOpus. It all started with, what is the meaning of life.

GPT4 has a good view of it.

interview.sh  GPTFunctionAgentWithMemoryAndBrowsingCreative "Speculate - what is the meaning of life?"  20 ChatGPT4TurboToolAgent  GeminiStraight PerplexityFunction MistralToolAgent GrokLPUFunctionAppLLama270b AnthropicOpusXMLFunctionAgent

1. execute the shell script using the configured, 'GPTFunctionAgentWithMemoryAndBrowsingCreative' flowise workflow as the moderator.
2. Start with the question, "Speculate - what is the meaning of life?"
3. 20 rounds of interview in this run
4. Ask the question to the configured flowise workflows: ChatGPT4TurboToolAgent, GeminiStraight, PerplexityFunction, MistralToolAgent GrokLPUFunctionAppLLama270b, AnthropicOpusXMLFunctionAgent
5. The moderator asks each of the configured Flowise AI workflows the question.
6. Each answer is collected and then after all have been asked the moderator summarizes the answers.
7. A new question is asked diving deeper into something from the summary.
8. Each round is appended to the output file.
9. It continues to run until the loop quantifier is reached. In this case 20 rounds.

interview_log.txt
interview.script.txt

Simple shell script to create long form blogs

#!/bin/bash

# Define an array of prompts
prompts=(
    "Introduction: My personal story of how I discovered the Law of Attraction and how it transformed my life"
    "What is the Law of Attraction? Explain the basic principles and how it works"
    "The science behind the Law of Attraction: Discuss research and evidence that supports it"
    "How to get started with the Law of Attraction: Practical tips for beginners to begin implementing it"
    "The importance of gratitude and positive thinking in manifesting your desires"
    "Overcoming limiting beliefs and negative thought patterns that block your manifestations"
    "How to get clear on what you really want: The power of setting intentions and being specific"
    "Visualization techniques to help you align with and attract what you desire"
    "Success stories: Examples of people who have used the Law of Attraction to manifest amazing things"
    "Conclusion: Encouragement and inspiration for readers to test out the Law of Attraction in their own lives"
)

# Iterate over the array of prompts
for prompt in "${prompts[@]}"; do
    # Execute the command with the current prompt
    echo "Running $prompt"
    tail -n 30 ~/Downloads/outputpost.md | fluent GPTFunctionAgentWithMemoryAndBrowsingCreative "You are writing a long blog post about the Law of attraction.  This section is: $prompt.  Do not offer unnatural introductions or lead ins, do not conclude or summarize the sections unnecessarily. You have access to the previous sections 20 lines in the context." >> ~/Downloads/outputpost.md
    
    # Wait for the command to complete before moving to the next one
    wait
done

echo "All prompts have been processed."

The attached file was generated from that script. It's just an example, so much is possible and this is just me playing around.

1. This file reads each line from the prompt collection in one at a time.
2. The prompts are executed with fluent getting the last 30 lines of the output as context to help smooth transitions.
3. The word output is appended to the output file.
4. The completed 5247 word document is attached below.

outputpost.md

The Help Page

Fluent 0.2.5

Interacts with Flowise AI workflows

USAGE:
    fluent [FLAGS] [OPTIONS] <name> [ARGS]

FLAGS:
        --generate-completions    
    -h, --help                    Prints help information
    -m, --markdown                
    -p, --parse-structured        
    -q, --quiet                   Decrease verbosity (-q, -qq, -qqq, etc.)
    -V, --version                 Prints version information
    -v, --verbose                 Increase verbosity (-v, -vv, -vvv, etc.)

OPTIONS:
        --dump-completions <shell>
            Write a completion definition for the specified shell to stdout (bash, zsh, etc.)

    -i, --image-folder-path <image-folder-path>                        
    -o, --outline-path <outline-path-flag>                             
    -s, --system-prompt-override <system-prompt-override>              
    -f, --system-prompt-override-file <system-prompt-override-file>    
    -t, --timestamp <resolution>
            Display timestamps on log messages (sec, ms, ns, none)


ARGS:


    <name>                       
    <question>                   
    <outline-path-positional>    
    <context>   

Summary

Fluent opens up the amazing FlowiseAI ecosystem by integrating it with the command line (linux/macos only right now). It's mind blowing what's possible and I'm only scratching the surface.

Programmed in Rust. It's a working build with poor tests, thats the next step.

[HenryHengZJ]

wow this is mind blowing! thank you so much for putting the effort into this! is there anyway i can get access to test it out? Would be even more powerful if we come up with a way to let user generate the flow without having to go to UI to construct

[njfio]

Yes, please! Absolutely! Invite sent.

Just a note for clarity, fluent isn't a (at present) a builder of the flows from cli. Quite the contrary. It's a wrapper that allows simple execution of the flows from CLI with pipeline and scripting support. Flowise is amazing at building the workflows. Fluent is pretty damn incredible at executing them in meaningful and unique ways that previously were very challenging to do with any of the AI systems I've used. Fluent at present is nothing without the magic of Flowise which I suppose itself is not much without the magic of LangChain friends.

I'm just starting to grasp what's possible when you can easily chain language models to each other or to other systems and commands.

That said, the tests are completely worthless at the moment but all features listed are working off the main branch.

I was playing with some scripts . Thought I would share this one with you. It uses Flowise, fluent, Make.com and Mem to create a little poem picture book. This one is based on FlowiseAI. I had to include the output as a PDF because the images are temporarily public after generation and I haven't gone through the effort to build the code to swap the images with a permanent store yet.

FlowiseAI Story_smallest.pdf

Shell script and it's config file.
process_poem copy.sh.txt
poem_process_config.json

Chatflows
MistralToolAgent Chatflow.json
Realvisxl Chatflow.json
Sonnet Function Chatflow.json

Please reach out if you have any questions or or would like to collaborate on the vision.

Thank you so much!
Nick

Interesting note, I used the same script structure that made that poem output to review the source code of the project and write the readme.md. It's far from perfect but a pretty good start. The magic of that script is that it parses input responses or file content as chunks split on blank new lines and feeds that chunk into fluent as context to the request sent to the workflow. This technique seems to work for outlines and poems pretty well. Again, it's just me playing around with the power of flowise and fluent.

One more note about the generation of flows without the ui, the configuration json allows full support (I think) for the override functionality within Flowise payloads and unwraps it directly. This in theory allows for more flow configuration diversity invoked through fluent than just the configured workflows in Flowise.

[HenryHengZJ]

I initially thought the pictures were hand drawn, they looked so cute! Do you happen to have the chatflows as well? One side question, what's the additional features/benefits that user can get as compared to using the API with overrideConfig

[njfio]

I updated the original reply with the additional files.

The biggest feature benefit to using fluent to me is the ability to inject any Flowise LLM workflow into the power of the linux shell, pipeline, and scripting capabilities.

Fluent doesn't change anything about Flowise or offer really anything your API doesn't. I've just made calling the workflows simplistic from the cli. Simplistic enough it's becoming my go to tool to work with any llm and all things linux shell related.

Sometimes I want a chat interface and sometimes I want so much more power and to be able to send files, text, and data between flows and linux processes that just aren't easy to do over and over with the Flowise CLI unless I make individual scripts for it. Fluent is my abstraction of what I was using over and over again from the Flowise API.

I love the concept of small purposeful Flowise workflows that output with expected behaviors and I love the concept of being able to chain them together easily by piping inputs and outputs into the workflows.

I find the smarter we ask AI systems to be and the more freedom we give them in solving problems the higher the cost is in running them. On multiple occasions I've blown through $500 in OpenAI charges on a weekend working with AutoGPT. I always found the results a bit less than what I really wanted and really desired a way to control what was going on with it more and to use those concepts with more than just GPT4.

Last night, I wrote a quick functional test for fluent so I could easily validate all the flags, while I am working on building the unit testing and restructuring the code. These are all included in the source and in the zip below. It runs through 19 different tests to validate all flags are processing as expected with proper input and expected output. I will likely extend this functionality to export to Airtable for analysis and probably a pipeline to evaluate the output by another workflow as well.

functional_test.zip

Now, I could expand that to take an array of flows and test all of my Flowise workflows in parallel in the background while I continue working and then evaluate it. Below is an example to run through all my currently configured Flowise workflows in parallel asking the same question to each about the source code of the project. It's not that this isn't possible with the Flowise API it's just the effort required to build it is pretty intense.

parallel-flow-example.sh.txt

Initially the addition of the overrideConfig in fluent was implemented so I could make inline adjustments to the system prompt. This expanded to including a file for the system prompt and on and on.

[amansoniamazatic]

Hi @njfio , can I also get access please?

[njfio]

Yes, delighted to provide it! Invite sent. Please reach out if you have any questions.

Thank you!
Nick

[amansoniamazatic]

Received, thank you.

[njfio]

While I found the proof of concept for Fluent to be incredibly useful the closer I got to releasing it the more I realized how much setup is involved just to get it running. Setting up fluent is very straightforward thought cumbersome but to be useful at all it requires flowise chatflows to be configured.

I work with engineers and devs every day and have been surprised by how many feel intimidated/overwhelmed by the ecosystem presented by Flowise. This isn't a problem with Flowise specifically. To someone just entering the AI and LLM landscape it's all a lot to understand. How X goes where and what works with what and all those questions. Even though, in my opinion, Flowise dumbs this down as simplistic as possible, it's still a lot to come to terms with for those just starting.

I also played with deploying Flowise on many different providers and each time I realized my configurations and keys needed to be restored, reconfigured, re-entered, etc. It was a lot, like a lot. I started to dread it. I also found that long-term nodes on chatflows get deprecated in favor of newer and better components which becomes a lot to deal with if you have a lot of chatflows configured.

So, I started a rewrite of the proof of concept to solve a bunch of these challenges ( for my use case which is all CLI based) and it's about 50% feature complete now. It has several enhancements that solve a variety of the challenges I experienced over my time working with Flowise and Fluent.

Number 1 new feature, integration of amber, https://github.com/fpco/amber to store the keys and any other aspect of the configuration desired. Amber is an amazing little utility and the end functionality in Fluent was unexpectedly awesome.

Amber encrypt secrets in a yaml file that is cryptographically secure and is unlocked with a private key stored as an env which allows it to securely provide credentials in pipelines. My current file's contents look like this:

file_format_version: 1
public_key: cce0cc20077c58ebe6422dfa87c0701c44bff83755922ed20bf4a7ec676f4a07
secrets:
- name: AMBER_ANOTHERWEBSERVICE_NJF
  sha256: 6b2b7498466801ea974e10a19de5c8c9deb5bc95ec98992a46a291d02a1f7b60
  cipher: 7d7725504623b21f90a0fada17d05ecaf0e1bc4f709baabc9f325dd4f1d0091c97d32a5aa77534e9ce0b678797418701474a018b7ff0535bc7cd3e8ccfcf57e4760c512ff6e3082c4e94999beabe148898b1a4fdcba26746b292455
- name: AMBER_FLUENT_ANTHROPIC_KEY_01
  sha256: 8f66c5f92b94d9228e9d32b30a29f3848fcd4ad9f4ef1ec4d153559be015c3e9
  cipher: da52538627a1a409458d749f43bf74138fd9aa9b0a8dc9e441ef15035aaae84f4b111cfa894ee9462231f90401177db64dc72c2c7dd8c38f79c36e61a10f592a4fcb1b973220c35854370ec57fdc62886d87398480cdd01aa75b7913712633147628606c9ad832b8b2e837937b7f6e70fd4e1615a23b3d72e5a26d2838c6e03728430032a4b8cbe16a3be07a98136cc191d3a56709a2b23efc76e06
- name: AMBER_FLUENT_GROQ_API_KEY_01
  sha256: 89f7aeec36590e47a2eab221b2381dd70343ae36b13e8e17a9f6dc62f1cb80be
  cipher: 6ff9c99f465f1354cc8a107d5cbbd1165be08dd1c0a77b67bbb22dd4866dcb2a2bb48622edeb0bfcc1f3b100b1a83871e6ea90dde44ee9b4909641af7bfd69790b45518884bb686cc22045441ba48d668d87b3893869bc25bf1c21777d2c132322349f387b8730b
- name: AMBER_FLUENT_MISTRAL_KEY_01
  sha256: 9731bc831969cfa36c45cbd7fb6b757c9cfbba91a96a64a847f9b6205d84fd2a
  cipher: 1c0bd5f0a66ee85221bbd179cf342b191105e36413a90d3740a6e5c5ef716939df7144ebb1a1bbb9eb068664c76c335bc455c6bfe6b3ebb7bc40fdcea97cb4e4ce46eef06d95f127ecafe3a985a0ca56
- name: AMBER_FLUENT_OPEN_API_KEY_01
  sha256: 50955d905232e8fb90d164e87fbf4a3fd49186a3220ef1b3197bf64725340483
  cipher: 04705361f04fa1e21d36b444ca36a8adfcf749564625cc3752bf3908a96aa00168683fe6855c9d14f20d082e3a9fa57e7493a1540c3932697c7242c8be5e0cbb4612b19b90933f6661e62eaa57349d54713f0d32e943d2ce6c6fdcdcaa350ba7ac7b0
- name: AMBER_FLUENT_SEARCHAPI_KEY_ID_01
  sha256: 3b378ff8071fe87a6311f10bcdd3f6171e1a0d08be1e4c4a70561ad9f2b24ec6
  cipher: 09aad10b270d089bbc5ec6ee160d217b65100c4e2bd0a8c06e16134efaa69868458fbd0b5b4559e43f6fb49c3bb691aabbc37c67ddf2d0257245b2068cc46139103b8b00b2d932c
- name: AMBER_FLUENT_SERPAPI_KEY_01
  sha256: fb632d090f352a7cddd40e2bdb25f72d2b1c9d5db075aacfde07ac395bf2a3de
  cipher: 72add13fc48436e451dd78d2dd2fa68fc5725edc89701db0a7a3db1073e16f5109ba2be9a99dd42ed51a66edd7caa7a600406c199df935fb880a64975faa4f3dd933f6fa471932250f7242a29d60191c096186c73cf2b9d42d7f747c9dd78a0ea9d987b7bb606b16821b71c45ee28a7
- name: AMBER_FLUENT_SESSION_ID_01
  sha256: d215bada471ed79a9ec9baf5e8d23a57c72aa195e37e81bea3050690340a2dd9
  cipher: 14d063487e9602ab97aae558ec5f2c2528ec667e5fbce9caca0c2586baf6cd466b49e90355fd19d67b19323cd37156679d1497684021c40ebc70d
- name: AMBER_FLUENT_ZEP_MEMORY_KEY_01
  sha256: 136bb9b51f44dc40238834475695f11c40970fccdeb4ab27766e6f7a405c32f1
  cipher: 7b5fc863ce222f5baae691a2dae429026f373e2f6729300902dca3ec147bb86ecea0115e4e0427e5ae50911ae032e7bc910c89d4a97b3807564b593b0d8dd1082eb8123a8c6d75b4e8cc4907d1baea92d1cc0098926a92573a759e0037f03fe41df653fad556301e81b734e2f1e7d2923ad75e1a5f1d818f7aeda805aa88b533b2e7b21d785d99182bf436936011c64e4063be5f609e0ba57c43d456cee39e3f3853b368cae7b2138cf75a87d08129cf99a7a3e1ef222f24bad87b992b9ee8e3c57721a48
- name: AMBER_LEONARDO_AI_KINO_XL_MODEL_ID
  sha256: 3b582f348b95a61ddbaa7ac3167a3f04f8e5e3178f0000f84ca8f0391f8aa1c1
  cipher: f48d253696fb03637462505e578cc8b8f759f24fe3b6a3c579237c1288c61d280dc457baf44dd39764aa195230bf748a0f824e81bdd00f918ab610128f235ba4591d69f7bdc4fd1347c43d54f8440b5447c29775
- name: AMBER_LOCAL_FLUENT_DEFAULT_KEY
  sha256: d6e8795e9d4a370659fce1ea86f03cb04efcc13b0927515faadc4e73b2a261f
  cipher: 0cd6f38d643190e0ab2e7ce74550d01f9e609d640358967114fe610e528e2f4645bc2c340e9a438e7e8c25b851ebc1a425e0cb74edab1c6a947a103a8fd5e1087fdcbd9140cef78bb79de0bd9ff1e5d399e4b1856b90b31e5926ced
- name: AMBER_MAKE_LEONARDO_IMAGE_POST
  sha256: d6e8795e9d4a370659fce1ea86f03cb04efcc13b0927515faadc4e73b2a261f
  cipher: 028fe2972087cc0211f0b6bfc1e6fd7365db93904d8b75580b3c2c9cdc695a09175a3ecbe2a601da51f5bbf12f48ade1724ed6286c24a125f3c5d3e6f9355956f291a361a4f13ad00d9ec66363db975f4929b6e5563cb30092f9dad

This is an example XML Agent with SerpAPI and Web Browsing tools in the chatflow. It's configured on a publicly available chatflow. The fluent configuration file is pasted below the image. Notice how the names in the amber file line up with values in the configuration file. For example: AMBER_FLUENT_SERPAPI_KEY_01

{
  "name": "SonnetXMLAgentAnowtherWebService",
  "protocol": "https",
  "hostname": "container-UNIQUECONTAINERID.containers.anotherwebservice.com",
  "port": 3000,
  "chat_id": "8ea46fa9-4aef-4184-a399-c588c576d148",
  "request_path": "/api/v1/prediction/",
  "sessionId": "",
  "bearer_token": "AMBER_ANOTHERWEBSERVICE_NJF",
  "overrideConfig": {
    "sessionId": "AMBER_FLUENT_SESSION_ID_01",
    "anthropicApiKey": "AMBER_FLUENT_ANTHROPIC_KEY_01",
    "stripNewLines": true,
    "modelName": {
      "chatAnthropic_0": "claude-3-sonnet-20240229",
      "chatOpenAI_0": "gpt-3.5-turbo-16k",
      "openAIEmbeddings_0": "text-embedding-3-small"
    },
    "openAIApiKey": {
      "openAIEmbeddings_0": "AMBER_FLUENT_OPEN_API_KEY_01",
      "chatOpenAI_0": "AMBER_FLUENT_OPEN_API_KEY_01"
    },
    "serpApiKey": "AMBER_FLUENT_SERPAPI_KEY_01",
    "SystemMessage": "You are a helpful assistant. Help the user answer any questions.\n\nYou have access to the following tools:\n\n{tools}\n\nIn order to use a tool, you can use <tool></tool> and <tool_input></tool_input> tags. You will then get back a response in the form <observation></observation>\nFor example, if you have a tool called 'search' that could run a google search, in order to search for the weather in SF you would respond:\n\n<tool>search</tool><tool_input>weather in SF</tool_input>\n<observation>64 degrees</observation>\n\nWhen you are done, respond with a final answer between <final_answer></final_answer>. For example:\n\n<final_answer>The weather in SF is 64 degrees</final_answer>\n\nBegin!\n\nPrevious Conversation:\n{chat_history}\n\nQuestion: {input}\n{agent_scratchpad}",
    "temperature": 0.9
  },
  "timeout_ms": 50000
}

What Fluent will now do on every execution is a dynamic replacement of any value in the configuration file that starts with AMBER_. It looks up the value in the amber store, exports the decrypted secret to temporary env keys, and uses the keys during execution to replace the entries in the configuration before submitting the payload to the chatflow, once executed the variables are removed from the environment table and the output is then processed.

This means one chatflow can be MANY different fluent configurations.

That same flow can have a different configuration using Haiku or Opus (or whatever) and turning the creativity down to 0 for example just by duplicating that configuration and altering the modelNames. This is incredibly useful for testing and evaluation of configuration settings and LLM outputs.

I'm adding Media (image/speech) Upload, Vector Upsert, and Document Loader functions soon. I will be releasing the new version for testing soon.

A few demo recordings of the original proof of concept.

• https://youtu.be/I0vGgAxdu0c
• https://youtu.be/zce3uvNEFNg
• https://youtu.be/zce3uvNEFNg
• https://youtu.be/zce3uvNEFNg

Nick

[2016bgeyer]

I'd love to get access to this! Not being able to interact more directly with the Flowise flows is a huge drawback for me. It would be one of the few things keeping me from using Flowise long-term instead of an alternative such as Prompt Flow from Microsoft

[njfio]

Yay! I'm grateful for your request. And good news! I rewrote it over the weekend improving every aspect and made it available this morning to a public repo.

https://github.com/njfio/fluent_cli

That said, I haven't written the full installation configuration docs yet, I'm doing it this afternoon and should have it posted by the morning.

It's straightforward enough but the integration with amber as a secure store does require some instruction. The good news is that once configured, it's version-able, and appears fully portable across system/os in my testing. I use the same file configuration file and amber store for both my mac and windows test environments.

I should also have a few public chatflows available and consumable with the included config.json file. The unique thing about the configs as I showed in a previous post, they are blank and will be filled in at runtime by fluent and amber.

I'll send an update to this thread when I have the install documents complete.

I'll close with this little gem that started working this morning. It's just a tiny free form example but combines some of the strengths of powershell with a GPT4 Flowise chatflow configured with search and browsing. Pretty awesome.

PS /Users/n> $fgm = fluent GPT4FunctionAgentWithMemoryAndBrowsing "Review the stats at https://feargreedmeter.com, create a json structure of all the details, output the json." -p | ConvertFrom-Json

⫸⫸⫸⫸⫸ 
Fluent:		GPT4FunctionAgentWithMemoryAndBrowsing
Processing:	Review the stats at https://feargreedmeter.com, create a json structure of all the details, output the json.
⫸⫸⫸⫸⫸ 

PS /Users/n> $fgm | fl -Force         
                                                                                                                                                               
FearGreedIndex : @{currentValue=50; previousClose=45; oneWeekAgo=40; oneMonthAgo=55; oneYearAgo=60; description=Neutral sentiment in the market.; factors=System.Object[]}

PS /Users/n> $fgm.FearGreedIndex | fl -Force

currentValue  : 50
previousClose : 45
oneWeekAgo    : 40
oneMonthAgo   : 55
oneYearAgo    : 60
description   : Neutral sentiment in the market.
factors       : {@{name=Market Momentum; value=60; description=Market is showing positive momentum.}, @{name=Stock Price Strength; value=40; description=Fewer stocks are hitting 
                new highs.}, @{name=Stock Price Breadth; value=50; description=Trading volumes in advancing stocks are balanced with declining ones.}, @{name=Put and Call Options; 
                value=55; description=Neutral sentiment in options market.}…}

PS /Users/n> $fgm.FearGreedIndex.factors | fl -Force

name        : Market Momentum
value       : 60
description : Market is showing positive momentum.

name        : Stock Price Strength
value       : 40
description : Fewer stocks are hitting new highs.

name        : Stock Price Breadth
value       : 50
description : Trading volumes in advancing stocks are balanced with declining ones.

name        : Put and Call Options
value       : 55
description : Neutral sentiment in options market.

name        : Junk Bond Demand
value       : 45
description : Investors are showing average interest in riskier bonds.

name        : Market Volatility
value       : 50
description : VIX (volatility index) is at typical levels.

name        : Safe Haven Demand
value       : 50
description : Normal demand for safe haven investments.

PS /Users/n> $fgm.FearGreedIndex.factors 

name                 value description
----                 ----- -----------
Market Momentum         60 Market is showing positive momentum.
Stock Price Strength    40 Fewer stocks are hitting new highs.
Stock Price Breadth     50 Trading volumes in advancing stocks are balanced with declining ones.
Put and Call Options    55 Neutral sentiment in options market.
Junk Bond Demand        45 Investors are showing average interest in riskier bonds.
Market Volatility       50 VIX (volatility index) is at typical levels.
Safe Haven Demand       50 Normal demand for safe haven investments.

https://youtu.be/Ntdf4eZ1BYI

[njfio]

The readme in the repository has installation instructions and basic usage details now.

These chatflow entries in the config.json will work once you go through the process of adding your api keys into the amber store. I'll add more in the coming days. You can override any setting you need to from the config.json file, hopefully with the examples I've provided it will be self-explanatory how to do that but if not please ask away. Better docs will be coming soon.

• GroqLLama370b8192AgentAnotherWebService
• GroqMixtral8x7bAgentAnotherWebService
• SonnetXMLAgentAnowtherWebService
• MistralLargeToolAgentAnowtherWebService

[njfio]

I wanted to provide an update. Today I released fluentCLI with support for media file upload, upsert with file, and upsert through other document loaders supported through FluentCLI and generalized chatflow configurations.

Let's look at an example.

This is a non-configured chatflow.

And this is an example configuration to call this chatflow from fluentCLI.

  {
    "name": "GPT4ToolAgentWithUpsertRepoCloud",
    "protocol": "https",
    "hostname": "9d81nz4o.rpcld.co",
    "port": 443,
    "chat_id": "931b1268-561d-494d-8f37-b5f5caac4466",
    "request_path": "/api/v1/prediction/",
    "upsert_path": "/api/v1/vector/upsert/",
    "sessionId": "",
    "bearer_token": "AMBER_REPO_CLOUD_FLUENT_DEMO_KEY",
    "overrideConfig": {
      "chainName": "GPT4ToolAgentWithUpsertRepoCloud",
      "sessionId": "AMBER_FLUENT_SESSION_ID_01",
      "openAIApiKey": "AMBER_FLUENT_OPENAI_API_KEY_01",
      "searchApiKey": "AMBER_FLUENT_SEARCHAPI_KEY_ID_01",
      "accessToken": "AMBER_FLUENT_GITHUB_PAT_KEY_01",
      "pineconeApiKey": "AMBER_FLUENT_PINECONE_API_KEY_01",
      "repoLink": "https://github.com/njfio/fluent_cli/",
      "branch":  "main",
      "recursive": true,
      "maxRetries" : 3,
      "language":{
        "codeTextSplitter_0": "rust"
      },
      "pineconeIndex": "fluentcli",
      "modelName": {
        "chatOpenAI_0": "gpt-4-turbo-preview",
        "openAIEmbeddings_1": "text-embedding-3-small"
      },
      "name": "fluentcli_retriever",
      "description": "fluentcli source code retriever",
      "returnSourceDocuments": true
    },
    "timeout_ms": 500000
  }

Great right but wouldn't it be even cooler to be able to change all of that inline too? This is the only command flag I allow full freeform inline override with currently but might add it to the rest. What does this mean?

fluent GPT4ToolAgentWithUpsertRepoCloud 'Hi, This is just upserting the FlowiseAI repo into the library, it is overriding the config.'   --upsert-no-upload '{ "repoLink": "https://github.com/njfio/Flowise"}'

Inline file upsert is also now supported.

target/release/fluent GPT4ToolAgentWithTextFileUpsertRepoCloud 'what is this ' --upsert-with-upload '/Users/n/Downloads/functional_test_log.txt'

Yesterday I improved the autocomplete functions for both Powershell and Bash. It is now possible to define a default flow in your profile and invoke it with fl or flps and everything after is taken as a request to sent. These commands still support stdin for pipelining context into the chatflow.

Piping the contents of a directory listing output into the chatflow. Not a fancy example but shows the pipeline.

These Powershell examples pipe the contents of an az vm list (lists all the VMs the logged-in account sees) into fluent to summarize the content.

Notice that the default options don't require quotations, everything after the fl or flps is taken as request input.

All of these examples should work with the included config.json in the source once you create the amber key vault.