From 5c4895356dc1a8d84ebdf5e2562a7a8d038fd890 Mon Sep 17 00:00:00 2001
From: JarbasAI <33701864+JarbasAl@users.noreply.github.com>
Date: Thu, 6 Jun 2024 19:23:23 +0100
Subject: [PATCH] Create file_formats.md

---
 docs/file_formats.md | 43 +++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 43 insertions(+)
 create mode 100644 docs/file_formats.md

diff --git a/docs/file_formats.md b/docs/file_formats.md
new file mode 100644
index 00000000..70d8f0c6
--- /dev/null
+++ b/docs/file_formats.md
@@ -0,0 +1,43 @@
+# OVOS File Formats Specification
+
+
+## Dialog Files
+
+- .dialog files
+- text files
+- each line is an alternative formulation of same sentence
+- supports "optional" syntax, eg `(word1|word2)` or `(optional word|)`
+- supports variables to be replaced at runtime, via curly braces syntax `{var_name}`
+
+## Intent files
+
+Example based intents (padatious style intents)
+
+contents are example sentences, Intent prediction models are trained on the content of these files
+
+- .intent files
+- text files
+- each line is an alternative formulation of same sentence
+- supports "optional" syntax, eg `(word1|word2)` or `(optional word|)`
+- supports variable capture, via curly braces syntax `{var_name}`
+
+## Voc files
+
+Keyword rule based intents (adapt style intents)
+
+contents are not a full sentence, but rather key words/phrases that must be present in the final sentence. Intents are defined by rules around .voc files (required, optional, at\_least\_one, excludes)
+
+- .voc files
+- text files
+- each line is an alternative formulation of same keyword
+- supports "optional" syntax, eg `(word1|word2)` or `(optional word|)`
+
+## Regex Files
+
+keyword intents do not allow variable capture, but they support regex expressions to capture keywords
+
+- .rx files
+- text files
+- each line is a regex expression, the capture group is the `variable_name`
+- `variable_name` becomes available as a keyword
+- complements .voc files