Merge pull request #62 from fractalego/development

Development

README.md

@@ -17,7 +17,7 @@ While it is ready to play with, it might not ready for production depending on y
 Installation
 ============
 
-In this version, WAFL is built to run as a two-part system.
+In this version, WAFL is a two-part system.
 Both can be installed on the same machine.
 
 ![The two parts of WAFL](images/two-parts.png)
@@ -65,29 +65,26 @@ Running WAFL
 This document contains a few examples of how to use the `wafl` CLI.
 There are four modes in which to run the system
 
-![The two parts of WAFL](images/wafl-commands.png)
-
-
-## wafl run-audio
+## $ wafl run-audio
 
 This is the main mode of operation. It will run the system in a loop, waiting for the user to speak a command.
 The activation word is the name defined in config.json.
 The default name is "computer", but you can change it to whatever you want.
 
 
-## wafl run-server
+## $ wafl run-server
 
 It runs a local web server that listens for HTTP requests on port 8889.
 The server will act as a chatbot, executing commands and returning the result as defined in the rules.
 
 
-## wafl run-cli
+## $ wafl run-cli
 
 This command works as for the run-server command, but it will listen for commands on the command line.
 It does not run a webserver and is useful for testing purposes.
 
 
-## wafl run-tests
+## $ wafl run-tests
 
 This command will run all the tests defined in the file testcases.txt.
 

changelog.txt

@@ -0,0 +1,5 @@
+### v 0.0. 70
+
+* Added the keyword RETRIEVE to get list of relevant items from the knowledge base
+* On-the-spot rule generation
+* mapping lists onto queries

documentation/build/html/_sources/directory_structure.rst.txt

@@ -39,5 +39,5 @@ Only the rules and facts that are included will be part of the inference.
 For example, the keyword `#using facts` within greetings/ (2) will not include the folder above it.
 Inference in a subfolder is limited the the rules and facts that are part of that folder or below it.
 
-For more information, you can have a look at the (still early) project in
+For more complete example, you can have a look at the (still early) project in
 `wafl_home <https://github.com/fractalego/wafl_home>`_.

documentation/build/html/_sources/examples.rst.txt

@@ -5,7 +5,4 @@ Examples
    :maxdepth: 2
 
    wafl_init
-   chitchat
-   rules
-   rules_and_backtracking
    directory_structure

documentation/build/html/_sources/index.rst.txt

@@ -3,8 +3,8 @@
    You can adapt this file completely to your liking, but it should at least
    contain the root `toctree` directive.
 
-Welcome to WAFL's documentation!
-================================
+Welcome to WAFL's 0.0.45 documentation!
+=======================================
 
 .. toctree::
    :maxdepth: 3
@@ -13,6 +13,8 @@ Welcome to WAFL's documentation!
    introduction
    installation
    running_WAFL
+   query_processing_pipeline
+   rules
    examples
    license
 

documentation/build/html/_sources/installation.rst.txt

@@ -9,21 +9,21 @@ Both can be installed on the same machine.
    :align: center
 
 Interface side
----------
+--------------
 
 The first part is local to your machine and needs to have access to a microphone and speaker.
 To install it, run the following commands:
 
 .. code-block:: bash
 
-    sudo apt-get install portaudio19-dev ffmpeg
-    pip install wafl
+    $ sudo apt-get install portaudio19-dev ffmpeg
+    $ pip install wafl
 
 After installing the requirements, you can initialize the interface by running the following command:
 
 .. code-block:: bash
 
-    wafl init
+    $ wafl init
 
 which creates a `config.json` file that you can edit to change the default settings.
 A standard rule file is also created as `wafl.rules`.
@@ -33,13 +33,16 @@ Please see the examples in the following chapters.
 LLM side (needs a GPU)
 ----------------------
 
-The second part is a server that runs on a machine with a public IP address.
-This last machine will need to have a GPU to run the Large Language Model at a convenient speed.
-This part can be run using a docker image by running the script
+The second part is a machine that runs on a machine accessible from the interface side.
+The initial configuration is for a local deployment of language models.
+No action is needed to run WAFL if you want to run it as a local instance.
+
+However, a multi-user setup will benefit for a dedicated server.
+In this case, a docker image can be used
 
 .. code-block:: bash
 
-    docker run -p8080:8080 --env NVIDIA_DISABLE_REQUIRE=1 --gpus all fractalego/wafl-llm:latest
+    $ docker run -p8080:8080 --env NVIDIA_DISABLE_REQUIRE=1 --gpus all fractalego/wafl-llm:latest
 
 
 The interface side has a `config.json` file that needs to be filled with the IP address of the LLM side.

documentation/build/html/_sources/introduction.rst.txt

@@ -1,8 +1,10 @@
 Introduction
 ============
 
-WAFL is a framework for home assistants.
-It is designed to combine Large Language Models and rules to create a predictable behavior.
+WAFL is a framework for personal agents.
+It integrates Large language models, speech recognition and text to speech.
+
+This framework combines Large Language Models and rules to create a predictable behavior.
 Specifically, instead of organising the work of an LLM into a chain of thoughts,
 WAFL intends to organise its behavior into inference trees.
 

documentation/build/html/_sources/rules.rst.txt

@@ -1,181 +1,11 @@
 Rules
 =====
 
-The file rules.wafl contains the rules used by the system.
-Each rule is in the following format
+Examples
+========
 
-.. code-block:: text
-
-    Trigger condition
-        action 1
-        action 2
-        action 3
-        ...
-
-Notice that the trigger condition has no indentation, while the actions are indented by any number spaces.
-Each action returns a true or false value.
-If that value is false, the rule stops executing and the next rule is triggered.
-A demo of the rules can be found in the repository `wafl_home <https://github.com/fractalego/wafl_home>`_.
-
-A rule ends when the next rule is encountered (a new trigger condition is found).
-The trigger condition can be a single fact. For example
-
-.. code-block:: text
-
-    This bot's name is "computer"
-
-There are two actors in the system: "the user" and "the bot".
-One simple rule example can be
-
-.. code-block:: text
-
-    The user asks what is this bot's name
-        SAY Hello, my name is Computer
-
-The rule above will be triggered when the user asks what is this bot's name.
-There are 7 types of actions:
-**SAY**,
-**REMEMBER**,
-**asking a question**,
-**generate a text**,
-**triggering of another rule**,
-**code execution**,
-**entailment**.
-
-
-SAY command
------------
-
-This command will make the bot say something.
-For example the rule above will make the bot say "Hello, my name is computer".
-
-REMEMBER command
-----------------
-
-This command will make the bot remember something.
-for example the rule below will make the bot remember the user's name.
-
-.. code-block:: text
-
-    The user says their name is John
-        REMEMBER The user's name is John
-
-Asking a question
------------------
-
-Typing a question (with or without question mark) will return a variable.
-This variable can be used later in the rule
-For example the rule below will make the bot ask the user's name.
-
-.. code-block:: text
-
-    The user says their name
-        name = what is the user's name?
-        REMEMBER The user's name is {name}
-
-Yes/No questions return a truth condition.
-For example by using the rule below, the bot will ask the user if they want to remember their name.
-
-.. code-block:: text
-
-    The user says their name
-        name = what is the user's name?
-        Do you want to remember the user's name?
-        REMEMBER The user's name is {name}
-
-If the user says "no", the rule will stop executing and the REMEMBER command will never be used
-
-
-Generate a text
-----------------
-
-A text can be generated in a similar fashion as when asking questions
-
-.. code-block:: text
-
-    The user says their name
-        name = what is the user's name?
-        italian_name = the italian version of {name} is
-        SAY The italian version of {name} is {italian_name}
-
-The text will be generated by the line "the italian version of {name}" according to the LLM model.
-The only difference with asking question is that the text on the right hand side of `=` is a statement
-and not a question.
-
-Triggering of another rule
---------------------------
-
-A rule can trigger another rule as follows
-
-.. code-block:: text
-
-    The user says their name
-        name = what is the user's name?
-        the name if the user is {name}
-
-    The name of the user is John
-        SAY Hello John!
-
-In this case the second rule is triggered if the user says their name is John.
-
-Code execution
---------------
-
-The code execution is done by using the python syntax.
-A function defined in the file `functions.py` can be called from the rule.
-
-
-For example, the file `rules.wafl` contains the following rule
-
-.. code-block:: text
-
-    The user says their name
-        name = what is the user's name?
-        greet({name})
-
-
-and the file `functions.py` contains the following function
-
-.. code-block:: python
-
-    def greet(name):
-        print("Hello", name)
-
-When the user says their name, the bot will greet the user by calling the function greet with the user's name as argument.
-However print() does not activate the SAY command.
-From the `functions.py` file, a rule can be triggered by using the syntax `"% ... %"`
-
-.. code-block:: python
-
-    def greet(name):
-        "% SAY Hello %"
-        f"% SAY your name is {name} %"
-
-The first line will make the bot say "Hello". The second line will make the bot say "your name is John" if the user's name is John.
-
-The syntax `"% ... %"`, can be used to trigger a rule, to generate a text, to ask a question, to remember something, or any other action available in the rules file.
-For example the prior function can be written as follows
-
-.. code-block:: python
-
-    def greet(name):
-        "% SAY Hello %"
-        "% SAY your name is {name} %"
-        date = "% what is the date today? %"
-        "% SAY today is {date} %"
-        while "% Do you want to continue? %":
-            "% SAY I am happy to continue %"
-
-Entailment
-----------
-
-The entailment is done by using the :- operator. if RHS entails LHS, then LSH :- RHS is true, otherwise it is false.
-For example the rule below will stop at the second line if the user's name is not John.
-
-.. code-block:: text
-
-    The user says their name
-        name = what is the user's name?
-        The user's name is John :- The user's name is {name}
-        SAY Your name is John!
+.. toctree::
+   :maxdepth: 2
 
+   writing_the_rules
+   rules_and_backtracking

documentation/build/html/_sources/running_WAFL.rst.txt

@@ -3,36 +3,30 @@ Running WAFL
 This document contains a few examples of how to use the `wafl` CLI.
 There are four modes in which to run the system
 
+$ wafl run-audio
+----------------
 
-.. image:: _static/wafl-commands.png
-   :alt: Basic CLI commands
-   :align: center
-
-
-wafl run-audio
---------------
-
-This is the main mode of operation. It will run the system in a loop, waiting for the user to speak a command.
+It will run the system in a loop, waiting for the user to speak a command.
 The activation word is the name defined in config.json.
 The default name is "computer", but you can change it to whatever you want.
 
 
-wafl run-server
----------------
+$ wafl run-server
+-----------------
 
 It runs a local web server that listens for HTTP requests on port 8889.
 The server will act as a chatbot, executing commands and returning the result as defined in the rules.
 
 
-wafl run-cli
-------------
+$ wafl run-cli
+--------------
 
 This command works as for the run-server command, but it will listen for commands on the command line.
 It does not run a webserver and is useful for testing purposes.
 
 
-wafl run-tests
---------------
+$ wafl run-tests
+----------------
 
 This command will run all the tests defined in the file testcases.txt.