refact

docs/docs/concepts.md

@@ -0,0 +1,91 @@
+# Weave concepts and lifecycle
+
+The Weave workflow can be broken into three major stages, organized as a loop:
+
+1. [Exploration](#exploration): Experiment with prompts, pipelines, and initial test cases.
+2. [Systematic Iteration](#systematic-iteration): Build evaluation datasets, compare models, and improve system performance.
+3. [Launch and Learn](#launch-and-learn): Deploy applications, collect feedback, and iteratively refine.
+
+Each stage includes specific substages that connect to core Weave features, as visualized in the diagram.
+
+![Weave usage lifecycle](../static/img/weave-cycle.png)
+
+## Exploration
+
+**Objective**: Early-stage experimentation to explore potential solutions and define tasks.
+
+### Experiment
+- **User Activities**: Experimenting with prompts and pipelines (e.g., Retrieval-Augmented Generation, or RAG).
+- **Related Weave Features**: 
+  - **App Tracing & Debugging**: 🟢 Lightweight tracing for quick insights.
+  - **LLM Playground**: 🔵/🚧 Integrated environment for replayable experimentation.
+  - **Cost Tracking**: 🟢 Understand and optimize resource usage.
+- **Outcome**: Insights into initial LLM behaviors and problem framing.
+
+### Fine-Tune
+- **User Activities**: Fine-tuning frontier models with domain-specific data.
+- **Related Weave Features**:
+  - **Prompt Management & Versioning**: 🔵/🚧 Track and refine prompt iterations.
+  - **Fine-Tune Tracking**: 💡 Monitor training improvements.
+- **Outcome**: Models tailored to specific domains or applications.
+
+---
+
+## Systematic Iteration
+
+**Objective**: Rigorous testing and evaluation to prepare for production.
+
+### Evaluate
+- **User Activities**: Building evaluation datasets and scoring methods.
+- **Related Weave Features**:
+  - **Evaluation Framework**: 🟢/📝 Code- and UI-based framework with intelligent caching.
+  - **Built-in Scorers**: 🔵 Automated scoring for datasets.
+- **Outcome**: Confidence in model accuracy and reliability.
+
+### Compare
+- **User Activities**: Comparing models and techniques.
+- **Related Weave Features**:
+  - **Model Management**: 🟢 Seamlessly organize and compare models.
+  - **Model Comparison Reports**: 🟢 Visualize differences in performance.
+  - **Leaderboards**: 🟢 Highlight top-performing approaches.
+- **Outcome**: Clear selection of the best-performing configurations.
+
+---
+
+## Launch and Learn
+
+**Objective**: Deploy LLM applications, collect feedback, and iterate.
+
+### Deploy
+- **User Activities**: Deploying the model/application.
+- **Related Weave Features**:
+  - **Production Tracing & Plotting**: 🟢 Monitor real-world behavior.
+  - **Guardrails and Alerts**: 🚧 Proactively identify issues in production.
+- **Outcome**: Applications ready for user interaction.
+
+### Enrich
+- **User Activities**: Collecting live user feedback and production data.
+- **Related Weave Features**:
+  - **Dataset Enrichment**: 🚧 Enhance evaluation datasets with production insights.
+  - **User Feedback Collection**: 🟢 Record and analyze interactions.
+- **Outcome**: Improved datasets and understanding of real-world use cases.
+
+### Fine-Tune
+- **User Activities**: Refining models based on production data.
+- **Related Weave Features**:
+  - **Fine-Tuning with Production Data**: 💡 Close the loop with improved performance.
+- **Outcome**: Enhanced model accuracy and responsiveness.
+
+---
+
+## Cross-Stage Foundations
+
+Weave includes foundational features that enhance every stage of the workflow:
+
+- **Multi-Client Support**: 🟢 Python, TypeScript, HTTP APIs, and more.
+- **Data Export**: 🟢 Export system data for external analysis.
+- **Saved Views**: 🚧 Share analytics and evaluations.
+- **Custom Mods**: 🚧 Build custom apps using Weave as a database.
+- **W&B Integration**: 🚧 Connect model development with evaluations and workflows.
+
+---

docs/docs/introduction.md

@@ -4,99 +4,7 @@ slug: /
 
 # Introduction to Weave
 
-Weave is a tool for exploring, evaluating, and deploying LLM-based applications. Designed for flexibility and scalability, Weave supports every stage of the LLM workflow, from initial experimentation to systematic iteration and deployment.
-
-The Weave workflow supports three major stages, organized as a loop:
-
-1. [Exploration](#exploration): Experiment with prompts, pipelines, and initial test cases.
-2. [Systematic Iteration](#systematic-iteration): Build evaluation datasets, compare models, and improve system performance.
-3. [Launch and Learn](#launch-and-learn): Deploy applications, collect feedback, and iteratively refine.
-
-Each stage includes specific substages that connect to core Weave features, as visualized in the diagram.
-
----
-
-![Weave usage lifecycle](../static/img/weave-cycle.png)
-
-## Exploration
-
-**Objective**: Early-stage experimentation to explore potential solutions and define tasks.
-
-### Experiment
-- **User Activities**: Experimenting with prompts and pipelines (e.g., Retrieval-Augmented Generation, or RAG).
-- **Related Weave Features**: 
-  - **App Tracing & Debugging**: 🟢 Lightweight tracing for quick insights.
-  - **LLM Playground**: 🔵/🚧 Integrated environment for replayable experimentation.
-  - **Cost Tracking**: 🟢 Understand and optimize resource usage.
-- **Outcome**: Insights into initial LLM behaviors and problem framing.
-
-### Fine-Tune
-- **User Activities**: Fine-tuning frontier models with domain-specific data.
-- **Related Weave Features**:
-  - **Prompt Management & Versioning**: 🔵/🚧 Track and refine prompt iterations.
-  - **Fine-Tune Tracking**: 💡 Monitor training improvements.
-- **Outcome**: Models tailored to specific domains or applications.
-
----
-
-## Systematic Iteration
-
-**Objective**: Rigorous testing and evaluation to prepare for production.
-
-### Evaluate
-- **User Activities**: Building evaluation datasets and scoring methods.
-- **Related Weave Features**:
-  - **Evaluation Framework**: 🟢/📝 Code- and UI-based framework with intelligent caching.
-  - **Built-in Scorers**: 🔵 Automated scoring for datasets.
-- **Outcome**: Confidence in model accuracy and reliability.
-
-### Compare
-- **User Activities**: Comparing models and techniques.
-- **Related Weave Features**:
-  - **Model Management**: 🟢 Seamlessly organize and compare models.
-  - **Model Comparison Reports**: 🟢 Visualize differences in performance.
-  - **Leaderboards**: 🟢 Highlight top-performing approaches.
-- **Outcome**: Clear selection of the best-performing configurations.
-
----
-
-## Launch and Learn
-
-**Objective**: Deploy LLM applications, collect feedback, and iterate.
-
-### Deploy
-- **User Activities**: Deploying the model/application.
-- **Related Weave Features**:
-  - **Production Tracing & Plotting**: 🟢 Monitor real-world behavior.
-  - **Guardrails and Alerts**: 🚧 Proactively identify issues in production.
-- **Outcome**: Applications ready for user interaction.
-
-### Enrich
-- **User Activities**: Collecting live user feedback and production data.
-- **Related Weave Features**:
-  - **Dataset Enrichment**: 🚧 Enhance evaluation datasets with production insights.
-  - **User Feedback Collection**: 🟢 Record and analyze interactions.
-- **Outcome**: Improved datasets and understanding of real-world use cases.
-
-### Fine-Tune
-- **User Activities**: Refining models based on production data.
-- **Related Weave Features**:
-  - **Fine-Tuning with Production Data**: 💡 Close the loop with improved performance.
-- **Outcome**: Enhanced model accuracy and responsiveness.
-
----
-
-## Cross-Stage Foundations
-
-Weave includes foundational features that enhance every stage of the workflow:
-
-- **Multi-Client Support**: 🟢 Python, TypeScript, HTTP APIs, and more.
-- **Data Export**: 🟢 Export system data for external analysis.
-- **Saved Views**: 🚧 Share analytics and evaluations.
-- **Custom Mods**: 🚧 Build custom apps using Weave as a database.
-- **W&B Integration**: 🚧 Connect model development with evaluations and workflows.
-
----
+Weave is a tool for exploring, evaluating, and deploying LLM-based applications, built by Weights & Biases (W&B). Designed for flexibility and scalability, Weave supports every stage of the LLM application development workflow, from initial experimentation to systematic iteration and deployment.
 
 ## Get started