Skip to content

Commit 074698f

Browse files
Shashikant86Shashikant86
committed
Update the docs
1 parent eaf92a8 commit 074698f

File tree

3 files changed

+611
-9
lines changed

3 files changed

+611
-9
lines changed

‎docs/examples/agents/gepa-integration.md‎

Lines changed: 6 additions & 6 deletions
Original file line numberDiff line numberDiff line change
@@ -135,20 +135,20 @@ spec:
135135
136136
```bash
137137
# Pull and set up agent
138-
super agent pull advanced_math_gepa
139-
super agent compile advanced_math_gepa
138+
super agent pull advanced-math-gepa
139+
super agent compile advanced-math-gepa
140140

141141
# Evaluate baseline
142-
super agent evaluate advanced_math_gepa
142+
super agent evaluate advanced-math-gepa
143143

144144
# Run GEPA optimization
145-
super agent optimize advanced_math_gepa
145+
super agent optimize advanced-math-gepa
146146

147147
# Measure improvement
148-
super agent evaluate advanced_math_gepa
148+
super agent evaluate advanced-math-gepa
149149

150150
# Test optimized agent
151-
super agent run advanced_math_gepa --goal "Solve x² + 3x - 4 = 0"
151+
super agent run advanced-math-gepa --goal "Solve x² + 3x - 4 = 0"
152152
```
153153

154154
### Memory-Efficient Configuration

‎docs/guides/dspy-optimizers.md‎

Lines changed: 276 additions & 3 deletions
Original file line numberDiff line numberDiff line change
@@ -471,15 +471,288 @@ Based on testing with llama3.1:8b on M4 Max (128GB):
471471
- [Memory Management](memory.md)
472472
- [Model Management](model-management.md)
473473

474+
## Quick Start Guide
475+
476+
### 🚀 Complete DSPy Optimizer Workflow
477+
478+
Here are ready-to-run commands for each DSPy optimizer. Each example includes pull, compile, optimize, and test steps.
479+
480+
#### GEPA - Graph Enhanced Prompting Algorithm
481+
**Best for**: Oracle-tier agents, complex reasoning, mathematical problems
482+
483+
```bash
484+
# Quick start with GEPA demo agent
485+
super agent pull gepa_demo
486+
super agent compile gepa_demo
487+
super agent optimize gepa_demo --timeout 300
488+
super agent evaluate gepa_demo
489+
super agent run gepa_demo --goal "Demonstrate reflective optimization capabilities"
490+
491+
# Advanced math with GEPA
492+
super agent pull advanced_math_gepa
493+
super agent compile advanced_math_gepa
494+
super agent optimize advanced_math_gepa --timeout 300
495+
super agent run advanced_math_gepa --goal "Solve x² + 5x - 6 = 0 showing all steps"
496+
```
497+
498+
#### SIMBA - Stochastic Introspective Mini-Batch Ascent
499+
**Best for**: Performance-critical applications, systematic improvement
500+
501+
```bash
502+
# Mathematics problems with SIMBA
503+
super agent pull simba_math
504+
super agent compile simba_math
505+
super agent optimize simba_math --timeout 300
506+
super agent evaluate simba_math
507+
super agent run simba_math --goal "Calculate the area of a circle with radius 7"
508+
509+
# General SIMBA optimization
510+
super agent pull simba_playbook
511+
super agent compile simba_playbook
512+
super agent optimize simba_playbook --timeout 300
513+
super agent run simba_playbook --goal "Optimize reasoning with mini-batch processing"
514+
```
515+
516+
#### MIPROv2 - Multi-step Instruction Prompt Optimization
517+
**Best for**: Instruction-following tasks, detailed reasoning
518+
519+
```bash
520+
# Mathematics with MIPROv2
521+
super agent pull miprov2_math
522+
super agent compile miprov2_math
523+
super agent optimize miprov2_math --timeout 300
524+
super agent evaluate miprov2_math
525+
super agent run miprov2_math --goal "Solve quadratic equation using multiple methods"
526+
527+
# General MIPROv2 optimization
528+
super agent pull miprov2_playbook
529+
super agent compile miprov2_playbook
530+
super agent optimize miprov2_playbook --timeout 300
531+
super agent run miprov2_playbook --goal "Demonstrate multi-step instruction optimization"
532+
```
533+
534+
#### BootstrapFewShot - Traditional Few-Shot Learning
535+
**Best for**: Getting started, reliable baselines, tool-calling agents
536+
537+
```bash
538+
# Mathematics with Bootstrap
539+
super agent pull bootstrap_math
540+
super agent compile bootstrap_math
541+
super agent optimize bootstrap_math --timeout 300
542+
super agent evaluate bootstrap_math
543+
super agent run bootstrap_math --goal "Solve algebraic equation with step-by-step explanation"
544+
545+
# General Bootstrap optimization
546+
super agent pull bootstrapfewshot_playbook
547+
super agent compile bootstrapfewshot_playbook
548+
super agent optimize bootstrapfewshot_playbook --timeout 300
549+
super agent run bootstrapfewshot_playbook --goal "Demonstrate traditional few-shot learning"
550+
```
551+
552+
#### BetterTogether - Ensemble Few-Shot Learning
553+
**Best for**: Robust performance, collaborative learning
554+
555+
```bash
556+
# Mathematics with BetterTogether
557+
super agent pull bettertogether_math
558+
super agent compile bettertogether_math
559+
super agent optimize bettertogether_math --timeout 300
560+
super agent evaluate bettertogether_math
561+
super agent run bettertogether_math --goal "Solve geometry problem using ensemble methods"
562+
563+
# General BetterTogether optimization
564+
super agent pull bettertogether_playbook
565+
super agent compile bettertogether_playbook
566+
super agent optimize bettertogether_playbook --timeout 300
567+
super agent run bettertogether_playbook --goal "Demonstrate ensemble optimization"
568+
```
569+
570+
#### COPRO - Collaborative Prompt Optimization
571+
**Best for**: Search-based improvement (requires special setup)
572+
573+
```bash
574+
# Set required environment variable
575+
export LITELLM_DROP_PARAMS=true
576+
577+
# Mathematics with COPRO
578+
super agent pull copro_math
579+
super agent compile copro_math
580+
super agent optimize copro_math --timeout 300
581+
super agent evaluate copro_math
582+
super agent run copro_math --goal "Solve calculus problem with collaborative optimization"
583+
584+
# General COPRO optimization
585+
super agent pull copro_playbook
586+
super agent compile copro_playbook
587+
super agent optimize copro_playbook --timeout 300
588+
super agent run copro_playbook --goal "Demonstrate collaborative prompt optimization"
589+
```
590+
591+
#### KNNFewShot - K-Nearest Neighbor Learning
592+
**Best for**: Pattern recognition, similarity-based learning
593+
594+
```bash
595+
# Mathematics with KNN
596+
super agent pull knn_math
597+
super agent compile knn_math
598+
super agent optimize knn_math --timeout 300
599+
super agent evaluate knn_math
600+
super agent run knn_math --goal "Solve trigonometry using pattern recognition"
601+
602+
# General KNN optimization
603+
super agent pull knnfewshot_playbook
604+
super agent compile knnfewshot_playbook
605+
super agent optimize knnfewshot_playbook --timeout 300
606+
super agent run knnfewshot_playbook --goal "Demonstrate K-nearest neighbor optimization"
607+
```
608+
609+
#### LabeledFewShot - Traditional Labeled Learning
610+
**Best for**: Small datasets, simple scenarios
611+
612+
```bash
613+
# Traditional labeled learning
614+
super agent pull labeledfewshot_playbook
615+
super agent compile labeledfewshot_playbook
616+
super agent optimize labeledfewshot_playbook --timeout 300
617+
super agent evaluate labeledfewshot_playbook
618+
super agent run labeledfewshot_playbook --goal "Demonstrate traditional labeled few-shot learning"
619+
```
620+
621+
---
622+
623+
### 📊 Comparison Workflow
624+
625+
Compare multiple optimizers on the same task:
626+
627+
```bash
628+
# Test all optimizers on math problems
629+
agents=("bootstrap_math" "bettertogether_math" "simba_math" "miprov2_math")
630+
631+
for agent in "${agents[@]}"; do
632+
echo "Testing $agent optimizer..."
633+
super agent pull $agent
634+
super agent compile $agent
635+
super agent evaluate $agent > baseline_$agent.txt
636+
super agent optimize $agent --timeout 300
637+
super agent evaluate $agent > optimized_$agent.txt
638+
echo "Results saved for $agent"
639+
done
640+
641+
# Compare results
642+
echo "Baseline vs Optimized Performance:"
643+
for agent in "${agents[@]}"; do
644+
echo "=== $agent ==="
645+
echo "Baseline:" && cat baseline_$agent.txt
646+
echo "Optimized:" && cat optimized_$agent.txt
647+
echo ""
648+
done
649+
```
650+
651+
---
652+
653+
### 🎯 Domain-Specific Quick Starts
654+
655+
#### For Mathematical Problem Solving
656+
```bash
657+
# Try different optimizers for math
658+
super agent pull advanced_math_gepa # GEPA for complex reasoning
659+
super agent pull simba_math # SIMBA for performance
660+
super agent pull miprov2_math # MIPROv2 for instruction clarity
661+
super agent pull bootstrap_math # Bootstrap for reliability
662+
663+
# Pick one and optimize
664+
super agent compile simba_math
665+
super agent optimize simba_math --timeout 300
666+
super agent run simba_math --goal "Find the integral of 2x³ + 3x² - x + 5"
667+
```
668+
669+
#### For General Purpose Optimization
670+
```bash
671+
# Try general-purpose optimizers
672+
super agent pull gepa_demo # GEPA demonstration
673+
super agent pull simba_playbook # SIMBA optimization
674+
super agent pull bettertogether_playbook # Ensemble methods
675+
super agent pull bootstrapfewshot_playbook # Traditional approach
676+
677+
# Pick one and test
678+
super agent compile bettertogether_playbook
679+
super agent optimize bettertogether_playbook --timeout 300
680+
super agent run bettertogether_playbook --goal "Demonstrate optimization capabilities"
681+
```
682+
683+
#### For Tool-Calling Agents (Genies Tier+)
684+
```bash
685+
# GEPA doesn't work with tool-calling agents
686+
# Use these optimizers instead:
687+
688+
super agent pull bootstrapfewshot_playbook # Recommended default
689+
super agent compile bootstrapfewshot_playbook
690+
super agent optimize bootstrapfewshot_playbook --timeout 300
691+
692+
# Alternative: SIMBA for complex tool interactions
693+
super agent pull simba_playbook
694+
super agent compile simba_playbook
695+
super agent optimize simba_playbook --timeout 300
696+
697+
# Alternative: BetterTogether for robust tool usage
698+
super agent pull bettertogether_playbook
699+
super agent compile bettertogether_playbook
700+
super agent optimize bettertogether_playbook --timeout 300
701+
```
702+
703+
---
704+
705+
## Available DSPy Optimizer Agents
706+
707+
SuperOptiX provides pre-configured agents demonstrating each DSPy optimizer:
708+
709+
### 🧮 Mathematics-Focused Agents
710+
711+
| Agent ID | Optimizer | Best For | Command |
712+
|----------|-----------|----------|---------|
713+
| `bootstrap_math` | BootstrapFewShot | Reliable baselines | `super agent pull bootstrap_math` |
714+
| `bettertogether_math` | BetterTogether | Robust performance | `super agent pull bettertogether_math` |
715+
| `simba_math` | SIMBA | Performance critical | `super agent pull simba_math` |
716+
| `miprov2_math` | MIPROv2 | Instruction clarity | `super agent pull miprov2_math` |
717+
| `copro_math` | COPRO | Search-based optimization | `super agent pull copro_math` |
718+
| `knn_math` | KNNFewShot | Pattern recognition | `super agent pull knn_math` |
719+
720+
### 🔧 General Purpose Agents
721+
722+
| Agent ID | Optimizer | Best For | Command |
723+
|----------|-----------|----------|---------|
724+
| `gepa_demo` | GEPA | Complex reasoning demo | `super agent pull gepa_demo` |
725+
| `bootstrapfewshot_playbook` | BootstrapFewShot | Traditional optimization | `super agent pull bootstrapfewshot_playbook` |
726+
| `bettertogether_playbook` | BetterTogether | Ensemble learning | `super agent pull bettertogether_playbook` |
727+
| `simba_playbook` | SIMBA | Advanced optimization | `super agent pull simba_playbook` |
728+
| `miprov2_playbook` | MIPROv2 | Multi-step instructions | `super agent pull miprov2_playbook` |
729+
| `copro_playbook` | COPRO | Collaborative optimization | `super agent pull copro_playbook` |
730+
| `knnfewshot_playbook` | KNNFewShot | Similarity-based learning | `super agent pull knnfewshot_playbook` |
731+
| `labeledfewshot_playbook` | LabeledFewShot | Traditional labeled learning | `super agent pull labeledfewshot_playbook` |
732+
733+
---
734+
474735
## Examples
475736

476737
Explore working examples in `/superoptix/agents/dspy_optimizers/`:
477738

478-
- `gepa_playbook.yaml` - GEPA configuration
739+
### Mathematical Problem Solving
740+
- `bootstrap_math_playbook.yaml` - Traditional few-shot for math
741+
- `bettertogether_math_playbook.yaml` - Ensemble learning for math
479742
- `simba_math_playbook.yaml` - SIMBA for mathematics
480743
- `miprov2_math_playbook.yaml` - MIPROv2 advanced prompting
481-
- `bootstrap_math_playbook.yaml` - Traditional few-shot
482-
- `bettertogether_math_playbook.yaml` - Ensemble learning
744+
- `copro_math_playbook.yaml` - Collaborative optimization for math
745+
- `knn_math_playbook.yaml` - K-nearest neighbor for math
746+
747+
### General Purpose Optimization
748+
- `gepa_playbook.yaml` - GEPA configuration and demonstration
749+
- `bootstrapfewshot_playbook.yaml` - Traditional few-shot learning
750+
- `bettertogether_playbook.yaml` - Ensemble learning methods
751+
- `simba_playbook.yaml` - SIMBA optimization
752+
- `miprov2_playbook.yaml` - Multi-step instruction optimization
753+
- `copro_playbook.yaml` - Collaborative prompt optimization
754+
- `knnfewshot_playbook.yaml` - K-nearest neighbor learning
755+
- `labeledfewshot_playbook.yaml` - Traditional labeled few-shot
483756

484757
## Contributing
485758

0 commit comments

Comments
 (0)