From 0cc24bc252a081ff95111acaed31bc4c2926becb Mon Sep 17 00:00:00 2001
From: Jeremy Frank <jfrank@summitinvest.tech>
Date: Sat, 26 Oct 2024 08:21:57 -0600
Subject: [PATCH] add README explaining agent workflow

---
 auto-content-creator/agents/README.md | 34 +++++++++++++++++++++++++++
 1 file changed, 34 insertions(+)
 create mode 100644 auto-content-creator/agents/README.md

diff --git a/auto-content-creator/agents/README.md b/auto-content-creator/agents/README.md
new file mode 100644
index 0000000..8c95118
--- /dev/null
+++ b/auto-content-creator/agents/README.md
@@ -0,0 +1,34 @@
+# Auto Content Creator Agent
+
+## Workflow Diagram
+
+```mermaid
+graph TD
+    START --> researchNode{Research Node}
+    researchNode -->|research needed| webSearch
+    researchNode -->|no research needed| generate
+    webSearch --> generate
+    generate -->|score < 9 & iterations < 10| reflect
+    generate -->|score >= 9 or iterations >= 10| END
+    generate -->|received feedback| processFeedback
+    reflect --> generate
+    processFeedback --> generate
+```
+
+### Workflow Steps Explanation:
+
+1. **Start**: The workflow begins when a content creation request is received
+2. **Research Node**: Dynamically decides if research is needed by analyzing:
+   - Current context (initial request vs feedback)
+   - Nature of the request/feedback
+   - Type of information needed
+3. **Generate**: Creates or updates content based on instructions/research/feedback
+4. **Decision Points**:
+   - If reflection score >= 9 or max iterations reached → END
+   - If human feedback received → Process Feedback
+   - Otherwise → Reflect
+5. **Reflect**: Evaluates content quality and provides a score
+6. **Process Feedback**: Incorporates human feedback to improve content
+7. **End**: Returns final content when quality threshold is met
+
+The workflow uses dynamic research decisions, performing research only when it would materially improve the response to the current request, whether that's an initial request or feedback iteration.