[feature] Allow tracking with bcc tools via openai function call.

[try-agaaain]

Description

This PR adds a command-line parameter -b(--bcc), when the user uses this parameter, GPTtrace will select the appropriate command tool from bcc-bpftools and set the appropriate parameters to complete the tracking task, like this:

$./GPTtrace.py -v -b "count kernel stack traces for submit_bio"
Run:  sudo stackcount-bpfcc submit_bio 
Tracing 1 functions for "submit_bio"... Hit Ctrl-C to end.
^C
  b'submit_bio'
  b'ext4_io_submit'
  b'ext4_bio_write_page'
  b'mpage_submit_page'
  b'mpage_process_page_bufs'
  b'mpage_prepare_extent_to_map'
  b'ext4_writepages'
  b'do_writepages'
  b'__writeback_single_inode'
  b'writeback_sb_inodes'
  b'__writeback_inodes_wb'
  b'wb_writeback'
  b'wb_workfn'
  b'process_one_work'
  b'worker_thread'
  b'kthread'
  b'ret_from_fork'
    2

[try-agaaain]

Problems that need to be solved

Here are some issues that I will fix later:

• bcc-bpftool has 100+ tools, but now only about 40 can be used because openai limits the maximum number of function calls to 64, and the number of tokens will exceed 4096 when all function calls are passed.
• The current project has a lot of code and needs to be re-structured to make the project code more clear.
• Some commands have positional parameters. When using these commands, the name of the position parameter cannot be added in front of the position parameter, e.g:
  stackcount-bpfcc --pid 112--pattern "submit" [wrong way]
  stackcount-bpfcc --pid 112 "submit" [right way]
  Here pattern is the location parameter.
• Documentation also needs to be updated.

[try-agaaain]

For the first problem, it is done in two steps:

• step1：Pass the description information of 117 commands to LLM, and let LLM determine the 20 commands that are most likely to be used to solve the user's problem; in this query, it is not necessary to use function call
• step2：Pass the 20 commands in step1 as function call, and let LLM decide which command to use to solve the user's problem.

For the third problem, I created a dictionary of positional parameters to determine which parameters of a given command are positional parameters, see https://github.com/try-agaaain/GPTtrace/blob/8b7c3eecdfeec04f0eb3eedba2e3526465374023/bcc_tools.py#L90

[yunwei37]

why not generate the function call dynamically at runtime? the advantages and disadvantages?

And,maybe you can try combined this with agents in langchain later?
https://docs.langchain.com/docs/components/agents/

[try-agaaain]

If I need to generate a function call dynamically, I first need to know which command to generate the function call, how should I determine this?

Dynamic generation has some drawbacks, as the function calls generated by LLM may not always be accurate. If we can predefine these function calls more accurately, it can reduce the occurrence of errors.

It doesn't seem to be more convenient to use Function Call in LangChain, but it's okay to change to LangChain.

[try-agaaain]

Does it possible to merge this PR first? It already includes a lot of updates (might be a bit messy). @yunwei37 @Officeyutong

I have added a -c option where users can use GPTtrace -c opensnoop-bpfcc "help me trace the open syscall in pid 123" to utilize opensnoop-bpfcc. For example:

```console
$./GPTtrace.py -c memleak-bpfcc "Trace allocations and display each individual allocator function call"
 Run:  sudo memleak-bpfcc --trace 
Attaching to kernel allocators, Ctrl+C to quit.
(b'Relay(35)', 402, 6, b'd...1', 20299.252425, b'alloc exited, size = 4096, result = ffff8881009cc000')
(b'Relay(35)', 402, 6, b'd...1', 20299.252425, b'free entered, address = ffff8881009cc000, size = 4096')
(b'Relay(35)', 402, 6, b'd...1', 20299.252426, b'free entered, address = 588a6f, size = 4096')
(b'Relay(35)', 402, 6, b'd...1', 20299.252427, b'alloc entered, size = 4096')
(b'Relay(35)', 402, 6, b'd...1', 20299.252427, b'alloc exited, size = 4096, result = ffff8881009cc000')
(b'Relay(35)', 402, 6, b'd...1', 20299.252428, b'free entered, address = ffff8881009cc000, size = 4096')
(b'sudo', 6938, 10, b'd...1', 20299.252437, b'alloc entered, size = 2048')
(b'sudo', 6938, 10, b'd...1', 20299.252439, b'alloc exited, size = 2048, result = ffff88822e845800')
(b'node', 410, 18, b'd...1', 20299.252455, b'alloc entered, size = 256')
(b'node', 410, 18, b'd...1', 20299.252457, b'alloc exited, size = 256, result = ffff8882e9b66400')
(b'node', 410, 18, b'd...1', 20299.252458, b'alloc entered, size = 2048')

For bcc tools, it looks up the corresponding function call from funcs.json. For other tools not defined in funcs.json, the LLM dynamically generates the function call.

In the next PR, I will attempt to accomplish the tracing task using a method similar to autogpt. #10

[Officeyutong]

Some suggestions:

• If this project will be maintained in a long term, it's better to write enough and meaningful typing hints for python codes. You can refer to https://typing.readthedocs.io/en/latest/source/best_practices.html or https://pybit.es/articles/code-better-with-type-hints-part-3/ or ChatGPT for more information

[try-agaaain]

Some suggestions:

• If this project will be maintained in a long term, it's better to write enough and meaningful typing hints for python codes. You can refer to https://typing.readthedocs.io/en/latest/source/best_practices.html or https://pybit.es/articles/code-better-with-type-hints-part-3/ or ChatGPT for more information

Great suggestion! In the last submission, I added documentation for each function. The descriptions in the documentation may not be detailed enough, but it will be improved in the future.