1
00:00:00,700 --> 00:00:04,430
Let's prepare a question to ask the agent.

2
00:00:05,620 --> 00:00:07,970
We will start with a programming -related

3
00:00:07,980 --> 00:00:12,210
question to see if the agent uses the
Python REPL tool.

4
00:00:14,970 --> 00:00:22,640
So, question equals, generate the first
20 numbers in the Fibonacci series.

5
00:00:31,290 --> 00:00:38,540
Now, we ll use the executor to invoke the
agent with the question, output equals,

6
00:00:39,370 --> 00:00:51,250
agentExecutor .invoke and the argument
will be a dictionary, the keys input and

7
00:00:51,260 --> 00:00:58,700
the value fromTemplate .format of Q
equals question.

8
00:00:59,690 --> 00:01:06,880
The argument Q is the parameter of the
template, this is the template, it s the

9
00:01:06,890 --> 00:01:12,980
question asked by the user, I am running
it, I am running the code.

10
00:01:18,390 --> 00:01:25,640
Take a look at the debugging information,
this is the chain of thugs, look how it

11
00:01:25,650 --> 00:01:32,540
chose the right tool Python REPL and
generated the first 20 numbers in the

12
00:01:32,550 --> 00:01:35,640
Fibonacci series very well.

13
00:01:36,450 --> 00:01:40,680
By the way, output is a dictionary with

14
00:01:40,690 --> 00:01:52,320
two keys input and output, output of
input contains the template and output of

15
00:01:52,330 --> 00:02:03,980
output contains the text response of the
agent, that s awesome, the agent nailed it.

16
00:02:05,470 --> 00:02:11,940
Next, let s try a current events question
to see if the agent searches the web

17
00:02:11,950 --> 00:02:22,460
using DuckDuckGo, the question will be,
Who is the current Prime Minister of the UK?

18
00:02:32,950 --> 00:02:40,560
I am running the code, it chose
DuckDuckGo search for this task, it needs

19
00:02:40,570 --> 00:02:55,790
up to date information, very well, this
is the correct answer, please ignore this

20
00:02:55,800 --> 00:03:06,590
warning for now, it has nothing to do
with Langchain, React or LLMs, let s ask

21
00:03:06,600 --> 00:03:13,570
another question, something related to
history so that it will choose the

22
00:03:13,580 --> 00:03:20,640
Wikipedia tool, tell me about Napoleon
Bonaparte early life and I am running the

23
00:03:20,650 --> 00:03:35,890
code, very well, it has chosen Wikipedia,
the agent nailed it choosing the right

24
00:03:35,900 --> 00:03:39,150
tool for each question and getting the
answer right.

25
00:03:40,040 --> 00:03:44,930
Now, let s take a look at the last
example to see if the chain of thoughts

26
00:03:44,940 --> 00:03:56,360
features is working as it should, in the
template here, I will add answer the

27
00:03:56,370 --> 00:04:08,970
following questions in German as best you can,
I am running this cell, now back to

28
00:04:08,980 --> 00:04:19,780
invoking the agent executor, I am running
this code, take a look here, the agent

29
00:04:19,790 --> 00:04:25,820
chain of thoughts is clear, to answer
this question in German, I will first

30
00:04:25,830 --> 00:04:31,160
gather information about Napoleon
Bonaparte early life in English and then

31
00:04:31,170 --> 00:04:40,120
translate it into German, very well, this
is the output translated into German,

32
00:04:40,370 --> 00:04:47,080
that s it folks, take your time to soak
this all in, it s the key to becoming a

33
00:04:47,090 --> 00:04:48,440
chain LLM wizard,