Skip to content
Open
Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
110 commits
Select commit Hold shift + click to select a range
0f6d065
dspy setup
Avngrstark62 Dec 15, 2025
2820c17
implemented golden dataset preparation
Avngrstark62 Dec 15, 2025
6a26179
implemented training pipeline
Avngrstark62 Dec 16, 2025
5f0b5b6
updated the training workflow in dspy to only use SIMBA and GEPA
Avngrstark62 Dec 27, 2025
32e8a21
added prompt storage and promote/rollback features
Avngrstark62 Jan 21, 2026
e9c402b
DSpy Adding more strategies
rajeshs-toast Dec 20, 2025
f55bf73
Restructuring the code for strategies
rajeshs-toast Dec 21, 2025
82a9d2d
Reducing postgres connections pool and reduce connection reuse
rajeshs-toast Dec 21, 2025
23a7dae
DSPY integration
rajeshs-toast Jan 2, 2026
593993c
update the bindu handler to make it compatible with dspy
Avngrstark62 Jan 22, 2026
3a1b5f7
minor import fix in train.py
Avngrstark62 Jan 22, 2026
670b1ab
added cli scripts to easily run dspy train and canary functions
Avngrstark62 Jan 22, 2026
1a419a5
Create dspy_docs.md
Avngrstark62 Jan 23, 2026
914781d
moved content of dspy/config.py to settings.py
Avngrstark62 Jan 23, 2026
5e0cb64
moved content of dspy/postgres.py to server/storage/postgres_storage.…
Avngrstark62 Jan 23, 2026
9074d10
refactored dspy/prompts.py for clean code and removed unneccesary dsp…
Avngrstark62 Jan 23, 2026
17cb7cd
add README.md in dspy directory
Avngrstark62 Jan 23, 2026
9dcc29d
added prompt_id to tasks table for on-the-spot metrics calculation in…
Avngrstark62 Jan 26, 2026
0162166
update dspy/README.md
Avngrstark62 Jan 26, 2026
6fb9b6d
added enable_dspy parameter in agent_config
Avngrstark62 Jan 26, 2026
3726b7c
fix did related issues due to dspy
Avngrstark62 Jan 26, 2026
d618cdc
fix issues
Avngrstark62 Jan 26, 2026
7a82102
fix issues
Avngrstark62 Jan 26, 2026
724d8c4
add test cases for dspy runtime part
Avngrstark62 Jan 26, 2026
63f5ace
add did argument to dspy cli
Avngrstark62 Jan 28, 2026
b68d497
add feedback metric for simba in dspy train cli
Avngrstark62 Jan 28, 2026
36a1509
add did argument to canary cli
Avngrstark62 Jan 28, 2026
19300a5
cleanup
Avngrstark62 Jan 28, 2026
8cb9751
add README.md for dspy
Avngrstark62 Jan 28, 2026
0f19abd
add unit tests for dspy
Avngrstark62 Jan 28, 2026
e735eef
fix dspy tests
Avngrstark62 Jan 28, 2026
cd498d0
add dspy section to readme files in the root in different languages
Avngrstark62 Jan 29, 2026
e61bbd0
minor change
Avngrstark62 Feb 8, 2026
2727444
remove redundant file
Avngrstark62 Feb 8, 2026
589ab4b
added new migration file and reverted the changes in previous file fo…
Avngrstark62 Feb 14, 2026
048c0ee
minor change
Avngrstark62 Feb 14, 2026
2e6ef74
added cli commands for postgres migrations
Avngrstark62 Feb 14, 2026
8846bf3
Refactor prompt storage to use local JSON file instead of PostgreSQL
google-labs-jules[bot] Feb 19, 2026
224541d
Refactor prompt storage to use local JSON file instead of PostgreSQL
google-labs-jules[bot] Feb 19, 2026
030e603
Fix the Readme.
Yuvraj-Dhepe Feb 22, 2026
0875779
Update examples to use prompt class.
Yuvraj-Dhepe Feb 22, 2026
1e97c26
Fix file level docstrings.
Yuvraj-Dhepe Feb 22, 2026
a1708e2
Fix function signatures and unit test to remove unused DID and storage.
Yuvraj-Dhepe Feb 22, 2026
db43989
dspy setup
Avngrstark62 Dec 15, 2025
16b72b9
implemented golden dataset preparation
Avngrstark62 Dec 15, 2025
d8e8fa5
implemented training pipeline
Avngrstark62 Dec 16, 2025
44ff885
updated the training workflow in dspy to only use SIMBA and GEPA
Avngrstark62 Dec 27, 2025
a6b0a6c
added prompt storage and promote/rollback features
Avngrstark62 Jan 21, 2026
a8e65d4
DSpy Adding more strategies
rajeshs-toast Dec 20, 2025
f8f4e1a
Restructuring the code for strategies
rajeshs-toast Dec 21, 2025
cec42aa
Reducing postgres connections pool and reduce connection reuse
rajeshs-toast Dec 21, 2025
7e385d0
update the bindu handler to make it compatible with dspy
Avngrstark62 Jan 22, 2026
ff394a2
Create dspy_docs.md
Avngrstark62 Jan 23, 2026
c69d4e9
moved content of dspy/config.py to settings.py
Avngrstark62 Jan 23, 2026
eabd248
moved content of dspy/postgres.py to server/storage/postgres_storage.…
Avngrstark62 Jan 23, 2026
cd521d6
refactored dspy/prompts.py for clean code and removed unneccesary dsp…
Avngrstark62 Jan 23, 2026
911b739
add README.md in dspy directory
Avngrstark62 Jan 23, 2026
c0c2e7f
added prompt_id to tasks table for on-the-spot metrics calculation in…
Avngrstark62 Jan 26, 2026
3a9b25b
fix did related issues due to dspy
Avngrstark62 Jan 26, 2026
d03924c
fix issues
Avngrstark62 Jan 26, 2026
85904f5
add test cases for dspy runtime part
Avngrstark62 Jan 26, 2026
33e8a1f
add did argument to dspy cli
Avngrstark62 Jan 28, 2026
85be888
add did argument to canary cli
Avngrstark62 Jan 28, 2026
d8f824d
cleanup
Avngrstark62 Jan 28, 2026
9024ffc
add unit tests for dspy
Avngrstark62 Jan 28, 2026
7ec3a1a
fix dspy tests
Avngrstark62 Jan 28, 2026
ee400b0
minor change
Avngrstark62 Feb 8, 2026
51db267
remove redundant file
Avngrstark62 Feb 8, 2026
83132d5
added new migration file and reverted the changes in previous file fo…
Avngrstark62 Feb 14, 2026
cd0bbd3
minor change
Avngrstark62 Feb 14, 2026
cf7c529
minor changes
Avngrstark62 Feb 15, 2026
8f27be7
major changes
Avngrstark62 Feb 15, 2026
ef1ce74
update canary logic
Avngrstark62 Feb 16, 2026
9276332
reset manifest_worker
Avngrstark62 Feb 16, 2026
e1d17e8
minor change
Avngrstark62 Feb 16, 2026
bd2b250
remove enable_dspy from various places
Avngrstark62 Feb 16, 2026
f9fb3d3
remove redundant things
Avngrstark62 Feb 16, 2026
510d506
remove redundant things
Avngrstark62 Feb 16, 2026
1a25c47
minor change
Avngrstark62 Feb 16, 2026
8cf853b
minor change
Avngrstark62 Feb 16, 2026
1b9f6e4
minor change
Avngrstark62 Feb 16, 2026
035c703
minor change
Avngrstark62 Feb 16, 2026
70414d8
minor change
Avngrstark62 Feb 16, 2026
af0439b
minor change
Avngrstark62 Feb 16, 2026
e76affb
Update train.py
Avngrstark62 Mar 3, 2026
58f9dd8
Update train.py
Avngrstark62 Mar 3, 2026
07d9175
minor change
Avngrstark62 Mar 3, 2026
dcb1826
minor change
Avngrstark62 Mar 3, 2026
452fd5d
minor change
Avngrstark62 Mar 3, 2026
9850dea
minor change
Avngrstark62 Mar 3, 2026
cadb5a4
minor change
Avngrstark62 Mar 3, 2026
e13e44d
minor change
Avngrstark62 Mar 3, 2026
176befd
move bindu/dspy/cli to bindu/cli, fix minor issues in dspy directory
Avngrstark62 Mar 4, 2026
cb8df38
fix compatibility issues in prompt related files
Avngrstark62 Mar 4, 2026
6c4fcfe
update examples/beginner/agno_example.py to use prompt router
Avngrstark62 Mar 4, 2026
1149edc
add an agno example for live prompt routing for dspy
Avngrstark62 Mar 4, 2026
6775368
minor fixes
Avngrstark62 Mar 4, 2026
85cd740
canary working
Avngrstark62 Mar 4, 2026
777500b
fix dspy training pipeline
Avngrstark62 Mar 14, 2026
a651856
sync examples
Avngrstark62 Mar 14, 2026
9a07df4
sync examples
Avngrstark62 Mar 14, 2026
e335c76
sync examples
Avngrstark62 Mar 14, 2026
5a5188e
sync examples
Avngrstark62 Mar 14, 2026
12404b4
sync examples
Avngrstark62 Mar 14, 2026
b25d214
minor change
Avngrstark62 Mar 14, 2026
d56b622
sync english readme
Avngrstark62 Mar 14, 2026
8eef224
update main english readme
Avngrstark62 Mar 14, 2026
68e5d93
reset other readmes
Avngrstark62 Mar 14, 2026
cf07e4a
update readme's of other languages to add dspy
Avngrstark62 Mar 14, 2026
6b29d8e
add test cases for dspy
Avngrstark62 Mar 14, 2026
File filter

Filter by extension

Filter by extension


Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
44 changes: 43 additions & 1 deletion README.bn.md
Original file line number Diff line number Diff line change
Expand Up @@ -824,7 +824,49 @@ Bindu **ফ্রেমওয়ার্ক-অজ্ঞেয়বাদী**

<br/>

## 🧪 টেস্টিং
## � DSPy প্রম্পট অপটিমাইজেশন

Bindu **DSPy** সংহত করে, একটি ফ্রেমওয়ার্ক যা LLM প্রম্পটগুলি প্রোগ্রামেটিকভাবে অপটিমাইজ করে এজেন্টের পারফরম্যান্স উন্নত করতে। ম্যানুয়ালি প্রম্পট তৈরি করার পরিবর্তে, DSPy স্বয়ংক্রিয়ভাবে বাস্তব এজেন্ট প্রতিক্রিয়া ব্যবহার করে উন্নত সংস্করণ তৈরি এবং যাচাই করে।

### এটি কীভাবে কাজ করে

DSPy তিনটি মূল ক্ষমতা প্রদান করে:

1. **🎯 অপটিমাইজড প্রম্পটগুলি প্রশিক্ষণ করুন** - এজেন্ট ইন্টারঅ্যাকশন বিশ্লেষণ করুন এবং ভাল প্রম্পট তৈরি করুন
```bash
bindu train --did agent_did
```
এটি A/B পরীক্ষার রূপান্তর তৈরি করে: একটি "সক্রিয়" প্রম্পট (আপনার বর্তমান সংস্করণ) এবং একটি "প্রার্থী" (অপটিমাইজড সংস্করণ) কনফিগারযোগ্য ট্রাফিক বিভাজন সহ।

2. **📊 ক্যানারি রোলআউট** - উন্নত প্রম্পটের দিকে ধীরে ধীরে ট্রাফিক স্থানান্তর করুন
```bash
bindu canary --did agent_did
```

3. **🔄 লাইভ প্রম্পট রাউটিং** - প্রতিটি অনুরোধের জন্য গতিশীলভাবে সঠিক প্রম্পট সংস্করণ সেবা করুন
```python
from bindu.dspy.prompt_router import route_prompt

async def handler(messages: list[dict[str, str]]):
agent.instructions = await route_prompt(initial_prompt=agent.instructions)
return agent.run(input=messages)
```

### দ্রুত শুরু উদাহরণ

একটি Agno এজেন্টের সাথে সম্পূর্ণ কাজের উদাহরণের জন্য [examples/agno_dspy_example.py](examples/agno_dspy_example.py) দেখুন।

### DSPy প্রশিক্ষণের জন্য পূর্বশর্ত

- আপনার `.env` ফাইলে `OPENAI_API_KEY` সেট করুন
- এজেন্ট কাজের ইতিহাস এবং প্রতিক্রিয়া রেটিং সহ PostgreSQL কনফিগার করুন
- নিশ্চিত করুন যে এজেন্টগুলি প্রতিক্রিয়া স্কোর (0-1 বা 1-5 রেটিং স্কেল) সহ মিথস্ক্রিয়া লগ করছে

---

<br/>

## �🧪 টেস্টিং

Bindu **70%+ টেস্ট কভারেজ** বজায় রাখে:

Expand Down
44 changes: 43 additions & 1 deletion README.de.md
Original file line number Diff line number Diff line change
Expand Up @@ -1186,7 +1186,49 @@ Möchtest du Integration mit deinem Lieblings-Framework? [Lass es uns auf Discor

<br/>

## 🧪 Testing
## � DSPy-Prompt-Optimierung

Bindu integriert **DSPy**, ein Framework zur programmatischen Optimierung von LLM-Prompts, um die Performance von Agenten zu verbessern. Anstatt Prompts manuell zu erstellen, generiert DSPy automatisch und validiert verbesserte Versionen basierend auf echtem Agenten-Feedback.

### Funktionsweise

DSPy bietet drei Hauptfunktionen:

1. **🎯 Trainiere optimierte Prompts** - Analysiere Agenten-Interaktionen und generiere bessere Prompts
```bash
bindu train --did agent_did
```
Dies erstellt A/B-Test-Varianten: einen "aktiven" Prompt (deine aktuelle Version) und einen "Kandidaten" (optimierte Version) mit konfigurierbarem Traffic-Split.

2. **📊 Canary Rollout** - Verschiebe den Traffic schrittweise zu dem verbesserten Prompt
```bash
bindu canary --did agent_did
```

3. **🔄 Live Prompt Routing** - Diene dynamisch die richtige Prompt-Version bei jeder Anfrage
```python
from bindu.dspy.prompt_router import route_prompt

async def handler(messages: list[dict[str, str]]):
agent.instructions = await route_prompt(initial_prompt=agent.instructions)
return agent.run(input=messages)
```

### Quick Start-Beispiel

Siehe [examples/agno_dspy_example.py](examples/agno_dspy_example.py) für ein komplettes funktionierendes Beispiel mit einem Agno-Agenten.

### Voraussetzungen für DSPy Training

- Setze `OPENAI_API_KEY` in deiner `.env`-Datei
- Habe PostgreSQL mit Agenten-Aufgabenhistorie und Feedback-Bewertungen konfiguriert
- Stelle sicher, dass Agenten Interaktionen mit Feedback-Scores protokollieren (0-1 oder 1-5 Bewertungsskala)

---

<br/>

## �🧪 Testing

Bindu hält **70%+ Test-Coverage**:

Expand Down
44 changes: 43 additions & 1 deletion README.es.md
Original file line number Diff line number Diff line change
Expand Up @@ -823,7 +823,49 @@ Bindu es **agnóstico al framework** y está probado con:

<br/>

## 🧪 Pruebas
## � Optimización de Prompts DSPy

Bindu integra **DSPy**, un framework para optimizar programáticamente los prompts de LLM y mejorar el rendimiento del agente. En lugar de crear manualmente prompts, DSPy genera y valida automáticamente versiones mejoradas utilizando el feedback real del agente.

### Cómo funciona

DSPy proporciona tres capacidades clave:

1. **🎯 Entrena prompts optimizados** - Analiza las interacciones del agente y genera mejores prompts
```bash
bindu train --did agent_did
```
Esto crea variantes de prueba A/B: un prompt "activo" (tu versión actual) y un "candidato" (versión optimizada) con división de tráfico configurable.

2. **📊 Canary Rollout** - Desplaza gradualmente el tráfico al prompt mejorado
```bash
bindu canary --did agent_did
```

3. **🔄 Enrutamiento dinámico de prompts** - Sirve dinámicamente la versión correcta del prompt para cada solicitud
```python
from bindu.dspy.prompt_router import route_prompt

async def handler(messages: list[dict[str, str]]):
agent.instructions = await route_prompt(initial_prompt=agent.instructions)
return agent.run(input=messages)
```

### Ejemplo de inicio rápido

Consulta [examples/agno_dspy_example.py](examples/agno_dspy_example.py) para obtener un ejemplo completo y funcional con un agente Agno.

### Requisitos previos para entrenar DSPy

- Establece `OPENAI_API_KEY` en tu archivo `.env`
- Configura PostgreSQL con historial de tareas de agentes y calificaciones de feedback
- Asegúrate de que los agentes registren interacciones con puntuaciones de feedback (escala de calificación 0-1 o 1-5)

---

<br/>

## �🧪 Pruebas

Bindu mantiene **70%+ de cobertura de pruebas**:

Expand Down
44 changes: 43 additions & 1 deletion README.fr.md
Original file line number Diff line number Diff line change
Expand Up @@ -823,7 +823,49 @@ Vous voulez une intégration avec votre framework préféré ? [Faites-le nous s

<br/>

## 🧪 Tests
## � Optimisation des invites DSPy

Bindu intègre **DSPy**, un framework pour optimiser programmatiquement les invites de LLM afin d'améliorer les performances des agents. Au lieu de créer manuellement les invites, DSPy génère et valide automatiquement les versions améliorées en utilisant les commentaires réels des agents.

### Comment ça fonctionne

DSPy offre trois capacités clés :

1. **🎯 Entraîner les invites optimisées** - Analyser les interactions des agents et générer de meilleures invites
```bash
bindu train --did agent_did
```
Cela crée des variantes de test A/B : une invite "active" (votre version actuelle) et une version "candidate" (optimisée) avec division du trafic configurable.

2. **📊 Lancement Canary** - Décaler progressivement le trafic vers l'invite améliorée
```bash
bindu canary --did agent_did
```

3. **🔄 Routage d'invites en direct** - Servir dynamiquement la bonne version d'invite à chaque demande
```python
from bindu.dspy.prompt_router import route_prompt

async def handler(messages: list[dict[str, str]]):
agent.instructions = await route_prompt(initial_prompt=agent.instructions)
return agent.run(input=messages)
```

### Exemple de démarrage rapide

Consultez [examples/agno_dspy_example.py](examples/agno_dspy_example.py) pour un exemple complet et fonctionnel avec un agent Agno.

### Conditions préalables à la formation DSPy

- Définissez `OPENAI_API_KEY` dans votre fichier `.env`
- Configurez PostgreSQL avec l'historique des tâches des agents et les évaluations de rétroaction
- Assurez-vous que les agents enregistrent les interactions avec des scores de rétroaction (échelle de notation 0-1 ou 1-5)

---

<br/>

## �🧪 Tests

Bindu maintient **70%+ de couverture de tests** :

Expand Down
44 changes: 43 additions & 1 deletion README.hi.md
Original file line number Diff line number Diff line change
Expand Up @@ -824,7 +824,49 @@ Bindu **फ्रेमवर्क-एग्नोस्टिक** है औ

<br/>

## 🧪 टेस्टिंग
## � DSPy प्रॉम्प्ट ऑप्टिमाइजेशन

Bindu **DSPy** को एकीकृत करता है, एक फ्रेमवर्क जो LLM प्रॉम्प्ट्स को प्रोग्रामेटिक रूप से ऑप्टिमाइज़ करने के लिए एजेंट प्रदर्शन में सुधार करता है। प्रॉम्प्ट्स को मैन्युअली क्राफ्ट करने की बजाय, DSPy वास्तविक एजेंट फीडबैक का उपयोग करके स्वचालित रूप से बेहतर संस्करण उत्पन्न और सत्यापित करता है।

### यह कैसे काम करता है

DSPy तीन मुख्य क्षमताएं प्रदान करता है:

1. **🎯 ऑप्टिमाइज़्ड प्रॉम्प्ट्स को प्रशिक्षित करें** - एजेंट इंटरैक्शन का विश्लेषण करें और बेहतर प्रॉम्प्ट्स उत्पन्न करें
```bash
bindu train --did agent_did
```
यह A/B टेस्ट वेरिएंट बनाता है: एक "सक्रिय" प्रॉम्प्ट (आपका वर्तमान संस्करण) और एक "उम्मीदवार" (ऑप्टिमाइज़्ड संस्करण) कॉन्फ़िगर करने योग्य ट्रैफिक विभाजन के साथ।

2. **📊 Canary Rollout** - बेहतर प्रॉम्प्ट के लिए क्रमिक रूप से ट्रैफिक स्थानांतरित करें
```bash
bindu canary --did agent_did
```

3. **🔄 लाइव प्रॉम्प्ट राउटिंग** - प्रत्येक अनुरोध के लिए गतिशील रूप से सही प्रॉम्प्ट संस्करण प्रदान करें
```python
from bindu.dspy.prompt_router import route_prompt

async def handler(messages: list[dict[str, str]]):
agent.instructions = await route_prompt(initial_prompt=agent.instructions)
return agent.run(input=messages)
```

### तेज़ शुरुआत के उदाहरण

[examples/agno_dspy_example.py](examples/agno_dspy_example.py) को एक Agno एजेंट के साथ एक पूर्ण कार्यप्रवाह उदाहरण के लिए देखें।

### DSPy प्रशिक्षण के लिए पूर्वापेक्षाएं

- अपनी `.env` फ़ाइल में `OPENAI_API_KEY` सेट करें
- एजेंट कार्य इतिहास और प्रतिक्रिया रेटिंग के साथ PostgreSQL को कॉन्फ़िगर करें
- सुनिश्चित करें कि एजेंट्स प्रतिक्रिया स्कोर (0-1 या 1-5 रेटिंग स्केल) के साथ इंटरैक्शन लॉग कर रहे हैं

---

<br/>

## �🧪 टेस्टिंग

Bindu **70%+ टेस्ट कवरेज** बनाए रखता है:

Expand Down
44 changes: 43 additions & 1 deletion README.md
Original file line number Diff line number Diff line change
Expand Up @@ -530,7 +530,49 @@ Want integration with your favorite framework? [Let us know on Discord](https://

<br/>

## 🧪 Testing
## � DSPy Prompt Optimization

Bindu integrates **DSPy**, a framework for programmatically optimizing LLM prompts to improve agent performance. Instead of manually crafting prompts, DSPy automatically generates and validates improved versions using real agent feedback.

### How It Works

DSPy provides three key capabilities:

1. **🎯 Train Optimized Prompts** - Analyze agent interactions and generate better prompts
```bash
bindu train --did agent_did
```
This creates A/B test variants: an "active" prompt (your current version) and a "candidate" (optimized version) with configurable traffic split.

2. **📊 Canary Rollout** - Gradually shift traffic to the improved prompt
```bash
bindu canary --did agent_did
```

3. **🔄 Live Prompt Routing** - Dynamically serve the right prompt version to each request
```python
from bindu.dspy.prompt_router import route_prompt

async def handler(messages: list[dict[str, str]]):
agent.instructions = await route_prompt(initial_prompt=agent.instructions)
return agent.run(input=messages)
```

### Quick Start Example

See [examples/agno_dspy_example.py](examples/agno_dspy_example.py) for a complete working example with an Agno agent.

### Prerequisites for DSPy Training

- Set `OPENAI_API_KEY` in your `.env` file
- Have PostgreSQL configured with agent task history and feedback ratings
- Ensure agents are logging interactions with feedback scores (0-1 or 1-5 rating scale)

---

<br/>

## �🧪 Testing

Bindu maintains **70%+ test coverage** (target: 80%+):

Expand Down
44 changes: 43 additions & 1 deletion README.nl.md
Original file line number Diff line number Diff line change
Expand Up @@ -823,7 +823,49 @@ Wil je integratie met je favoriete framework? [Laat het ons weten op Discord](ht

<br/>

## 🧪 Testen
## � DSPy Prompt Optimalisatie

Bindu integreert **DSPy**, een framework voor programmatische optimalisatie van LLM-prompts om de agentprestaties te verbeteren. In plaats van handmatig prompts te schrijven, genereert DSPy automatisch en valideert verbeterde versies met behulp van echte agent-feedback.

### Hoe het werkt

DSPy biedt drie essentiële mogelijkheden:

1. **🎯 Traingeoptimaliseerde prompts** - Analyseer agent-interacties en genereer betere prompts
```bash
bindu train --did agent_did
```
Dit creëert A/B-testvarianten: een "actieve" prompt (uw huidige versie) en een "kandidaat" (geoptimaliseerde versie) met configureerbare verkeersverdeling.

2. **📊 Canary Rollout** - Verschuif geleidelijk verkeer naar de verbeterde prompt
```bash
bindu canary --did agent_did
```

3. **🔄 Live Prompt Routing** - Dien dynamisch de juiste promptversie in voor elk verzoek
```python
from bindu.dspy.prompt_router import route_prompt

async def handler(messages: list[dict[str, str]]):
agent.instructions = await route_prompt(initial_prompt=agent.instructions)
return agent.run(input=messages)
```

### Snelstart Voorbeeld

Zie [examples/agno_dspy_example.py](examples/agno_dspy_example.py) voor een compleet werkend voorbeeld met een Agno-agent.

### Vereisten voor DSPy Training

- Stel `OPENAI_API_KEY` in uw `.env` bestand in
- Configureer PostgreSQL met agent-taakgeschiedenis en feedback-beoordelingen
- Zorg ervoor dat agents interacties met feedbackscores (0-1 of 1-5 beoordelingsschaal) registreren

---

<br/>

## �🧪 Testen

Bindu handhaaft **70%+ test coverage**:

Expand Down
Loading