feat(projects): add AgentProbe to featured portfolio (#3)

ReAct Agent Observatory with 8-type failure taxonomy and
multi-provider eval harness. Surfaces the agent-reliability
work alongside existing data and backend projects.

Author: soneeee22000

src/lib/data.ts

@@ -158,6 +158,16 @@ export const PROJECTS_META = [
     demo: "https://greenlens-api.azurewebsites.net",
     gh: "https://github.com/soneeee22000/GreenLens.dev",
   },
+  {
+    id: 12,
+    featured: true,
+    emoji: "\uD83D\uDD2C",
+    title: "AgentProbe",
+    tags: ["Python", "FastAPI", "Next.js 16", "PostgreSQL", "ReAct", "SSE"],
+    color: "#C9A96E",
+    demo: "#",
+    gh: "https://github.com/soneeee22000/AgentProbe",
+  },
 ];
 
 export const SKILLS: Skill[] = [

src/messages/de.json

@@ -88,7 +88,8 @@
     "7_desc": "Real-time urban mobility analytics on TimescaleDB + PostGIS + Uber H3. Spring Boot ingestion, Kafka stream processing, Server-Sent Events for live dashboards. Smart-city back-end blueprint.",
     "8_desc": "Reference implementation of the BCBS 239 risk-data-aggregation lakehouse pattern on Databricks + Delta Lake + Unity Catalog + dbt-databricks. MIT-licensed, synthetic data only — banking risk-data engineering portfolio piece.",
     "9_desc": "Serverless middleware enforcing responsible-AI compliance on LLM applications — PII detection, bias screening, hate-speech filtering, and RAG-powered policy enforcement on top of Azure OpenAI.",
-    "10_desc": "Cloud Carbon Intelligence platform estimating CO2e emissions from Azure infrastructure with AI-powered reduction recommendations. Clean Architecture, Azure AI Search semantic factor lookup, 88 automated tests. Built for EU CSRD Scope 3 compliance."
+    "10_desc": "Cloud Carbon Intelligence platform estimating CO2e emissions from Azure infrastructure with AI-powered reduction recommendations. Clean Architecture, Azure AI Search semantic factor lookup, 88 automated tests. Built for EU CSRD Scope 3 compliance.",
+    "11_desc": "From-scratch ReAct Agent Observatory — observe, debug, and benchmark LLM agents with a built-in 8-type failure taxonomy (hallucinated tools, malformed actions, context overflow, goal drift, …) and multi-provider eval harness (Groq, OpenAI, Anthropic, Google, Ollama). Composite scoring (answer + tools + efficiency + reliability), Clean Architecture FastAPI + PostgreSQL backend, Next.js 16 frontend with real-time SSE streaming. 50+ benchmark cases, 81 tests."
   },
   "skills": {
     "label": "04 — SKILLS",

src/messages/en.json

@@ -88,7 +88,8 @@
     "7_desc": "Real-time urban mobility analytics on TimescaleDB + PostGIS + Uber H3. Spring Boot ingestion, Kafka stream processing, Server-Sent Events for live dashboards. Smart-city back-end blueprint.",
     "8_desc": "Reference implementation of the BCBS 239 risk-data-aggregation lakehouse pattern on Databricks + Delta Lake + Unity Catalog + dbt-databricks. MIT-licensed, synthetic data only — banking risk-data engineering portfolio piece.",
     "9_desc": "Serverless middleware enforcing responsible-AI compliance on LLM applications — PII detection, bias screening, hate-speech filtering, and RAG-powered policy enforcement on top of Azure OpenAI.",
-    "10_desc": "Cloud Carbon Intelligence platform estimating CO2e emissions from Azure infrastructure with AI-powered reduction recommendations. Clean Architecture, Azure AI Search semantic factor lookup, 88 automated tests. Built for EU CSRD Scope 3 compliance."
+    "10_desc": "Cloud Carbon Intelligence platform estimating CO2e emissions from Azure infrastructure with AI-powered reduction recommendations. Clean Architecture, Azure AI Search semantic factor lookup, 88 automated tests. Built for EU CSRD Scope 3 compliance.",
+    "11_desc": "From-scratch ReAct Agent Observatory — observe, debug, and benchmark LLM agents with a built-in 8-type failure taxonomy (hallucinated tools, malformed actions, context overflow, goal drift, …) and multi-provider eval harness (Groq, OpenAI, Anthropic, Google, Ollama). Composite scoring (answer + tools + efficiency + reliability), Clean Architecture FastAPI + PostgreSQL backend, Next.js 16 frontend with real-time SSE streaming. 50+ benchmark cases, 81 tests."
   },
   "skills": {
     "label": "04 — SKILLS",

src/messages/es.json

@@ -88,7 +88,8 @@
     "7_desc": "Real-time urban mobility analytics on TimescaleDB + PostGIS + Uber H3. Spring Boot ingestion, Kafka stream processing, Server-Sent Events for live dashboards. Smart-city back-end blueprint.",
     "8_desc": "Reference implementation of the BCBS 239 risk-data-aggregation lakehouse pattern on Databricks + Delta Lake + Unity Catalog + dbt-databricks. MIT-licensed, synthetic data only — banking risk-data engineering portfolio piece.",
     "9_desc": "Serverless middleware enforcing responsible-AI compliance on LLM applications — PII detection, bias screening, hate-speech filtering, and RAG-powered policy enforcement on top of Azure OpenAI.",
-    "10_desc": "Cloud Carbon Intelligence platform estimating CO2e emissions from Azure infrastructure with AI-powered reduction recommendations. Clean Architecture, Azure AI Search semantic factor lookup, 88 automated tests. Built for EU CSRD Scope 3 compliance."
+    "10_desc": "Cloud Carbon Intelligence platform estimating CO2e emissions from Azure infrastructure with AI-powered reduction recommendations. Clean Architecture, Azure AI Search semantic factor lookup, 88 automated tests. Built for EU CSRD Scope 3 compliance.",
+    "11_desc": "From-scratch ReAct Agent Observatory — observe, debug, and benchmark LLM agents with a built-in 8-type failure taxonomy (hallucinated tools, malformed actions, context overflow, goal drift, …) and multi-provider eval harness (Groq, OpenAI, Anthropic, Google, Ollama). Composite scoring (answer + tools + efficiency + reliability), Clean Architecture FastAPI + PostgreSQL backend, Next.js 16 frontend with real-time SSE streaming. 50+ benchmark cases, 81 tests."
   },
   "skills": {
     "label": "04 — SKILLS",

src/messages/fr.json

@@ -88,7 +88,8 @@
     "7_desc": "Real-time urban mobility analytics on TimescaleDB + PostGIS + Uber H3. Spring Boot ingestion, Kafka stream processing, Server-Sent Events for live dashboards. Smart-city back-end blueprint.",
     "8_desc": "Reference implementation of the BCBS 239 risk-data-aggregation lakehouse pattern on Databricks + Delta Lake + Unity Catalog + dbt-databricks. MIT-licensed, synthetic data only — banking risk-data engineering portfolio piece.",
     "9_desc": "Serverless middleware enforcing responsible-AI compliance on LLM applications — PII detection, bias screening, hate-speech filtering, and RAG-powered policy enforcement on top of Azure OpenAI.",
-    "10_desc": "Cloud Carbon Intelligence platform estimating CO2e emissions from Azure infrastructure with AI-powered reduction recommendations. Clean Architecture, Azure AI Search semantic factor lookup, 88 automated tests. Built for EU CSRD Scope 3 compliance."
+    "10_desc": "Cloud Carbon Intelligence platform estimating CO2e emissions from Azure infrastructure with AI-powered reduction recommendations. Clean Architecture, Azure AI Search semantic factor lookup, 88 automated tests. Built for EU CSRD Scope 3 compliance.",
+    "11_desc": "From-scratch ReAct Agent Observatory — observe, debug, and benchmark LLM agents with a built-in 8-type failure taxonomy (hallucinated tools, malformed actions, context overflow, goal drift, …) and multi-provider eval harness (Groq, OpenAI, Anthropic, Google, Ollama). Composite scoring (answer + tools + efficiency + reliability), Clean Architecture FastAPI + PostgreSQL backend, Next.js 16 frontend with real-time SSE streaming. 50+ benchmark cases, 81 tests."
   },
   "skills": {
     "label": "04 — SKILLS",

src/messages/ja.json

@@ -88,7 +88,8 @@
     "7_desc": "Real-time urban mobility analytics on TimescaleDB + PostGIS + Uber H3. Spring Boot ingestion, Kafka stream processing, Server-Sent Events for live dashboards. Smart-city back-end blueprint.",
     "8_desc": "Reference implementation of the BCBS 239 risk-data-aggregation lakehouse pattern on Databricks + Delta Lake + Unity Catalog + dbt-databricks. MIT-licensed, synthetic data only — banking risk-data engineering portfolio piece.",
     "9_desc": "Serverless middleware enforcing responsible-AI compliance on LLM applications — PII detection, bias screening, hate-speech filtering, and RAG-powered policy enforcement on top of Azure OpenAI.",
-    "10_desc": "Cloud Carbon Intelligence platform estimating CO2e emissions from Azure infrastructure with AI-powered reduction recommendations. Clean Architecture, Azure AI Search semantic factor lookup, 88 automated tests. Built for EU CSRD Scope 3 compliance."
+    "10_desc": "Cloud Carbon Intelligence platform estimating CO2e emissions from Azure infrastructure with AI-powered reduction recommendations. Clean Architecture, Azure AI Search semantic factor lookup, 88 automated tests. Built for EU CSRD Scope 3 compliance.",
+    "11_desc": "From-scratch ReAct Agent Observatory — observe, debug, and benchmark LLM agents with a built-in 8-type failure taxonomy (hallucinated tools, malformed actions, context overflow, goal drift, …) and multi-provider eval harness (Groq, OpenAI, Anthropic, Google, Ollama). Composite scoring (answer + tools + efficiency + reliability), Clean Architecture FastAPI + PostgreSQL backend, Next.js 16 frontend with real-time SSE streaming. 50+ benchmark cases, 81 tests."
   },
   "skills": {
     "label": "04 — SKILLS",

src/messages/ko.json

@@ -88,7 +88,8 @@
     "7_desc": "Real-time urban mobility analytics on TimescaleDB + PostGIS + Uber H3. Spring Boot ingestion, Kafka stream processing, Server-Sent Events for live dashboards. Smart-city back-end blueprint.",
     "8_desc": "Reference implementation of the BCBS 239 risk-data-aggregation lakehouse pattern on Databricks + Delta Lake + Unity Catalog + dbt-databricks. MIT-licensed, synthetic data only — banking risk-data engineering portfolio piece.",
     "9_desc": "Serverless middleware enforcing responsible-AI compliance on LLM applications — PII detection, bias screening, hate-speech filtering, and RAG-powered policy enforcement on top of Azure OpenAI.",
-    "10_desc": "Cloud Carbon Intelligence platform estimating CO2e emissions from Azure infrastructure with AI-powered reduction recommendations. Clean Architecture, Azure AI Search semantic factor lookup, 88 automated tests. Built for EU CSRD Scope 3 compliance."
+    "10_desc": "Cloud Carbon Intelligence platform estimating CO2e emissions from Azure infrastructure with AI-powered reduction recommendations. Clean Architecture, Azure AI Search semantic factor lookup, 88 automated tests. Built for EU CSRD Scope 3 compliance.",
+    "11_desc": "From-scratch ReAct Agent Observatory — observe, debug, and benchmark LLM agents with a built-in 8-type failure taxonomy (hallucinated tools, malformed actions, context overflow, goal drift, …) and multi-provider eval harness (Groq, OpenAI, Anthropic, Google, Ollama). Composite scoring (answer + tools + efficiency + reliability), Clean Architecture FastAPI + PostgreSQL backend, Next.js 16 frontend with real-time SSE streaming. 50+ benchmark cases, 81 tests."
   },
   "skills": {
     "label": "04 — SKILLS",

src/messages/my.json

@@ -88,7 +88,8 @@
     "7_desc": "Real-time urban mobility analytics on TimescaleDB + PostGIS + Uber H3. Spring Boot ingestion, Kafka stream processing, Server-Sent Events for live dashboards. Smart-city back-end blueprint.",
     "8_desc": "Reference implementation of the BCBS 239 risk-data-aggregation lakehouse pattern on Databricks + Delta Lake + Unity Catalog + dbt-databricks. MIT-licensed, synthetic data only — banking risk-data engineering portfolio piece.",
     "9_desc": "Serverless middleware enforcing responsible-AI compliance on LLM applications — PII detection, bias screening, hate-speech filtering, and RAG-powered policy enforcement on top of Azure OpenAI.",
-    "10_desc": "Cloud Carbon Intelligence platform estimating CO2e emissions from Azure infrastructure with AI-powered reduction recommendations. Clean Architecture, Azure AI Search semantic factor lookup, 88 automated tests. Built for EU CSRD Scope 3 compliance."
+    "10_desc": "Cloud Carbon Intelligence platform estimating CO2e emissions from Azure infrastructure with AI-powered reduction recommendations. Clean Architecture, Azure AI Search semantic factor lookup, 88 automated tests. Built for EU CSRD Scope 3 compliance.",
+    "11_desc": "From-scratch ReAct Agent Observatory — observe, debug, and benchmark LLM agents with a built-in 8-type failure taxonomy (hallucinated tools, malformed actions, context overflow, goal drift, …) and multi-provider eval harness (Groq, OpenAI, Anthropic, Google, Ollama). Composite scoring (answer + tools + efficiency + reliability), Clean Architecture FastAPI + PostgreSQL backend, Next.js 16 frontend with real-time SSE streaming. 50+ benchmark cases, 81 tests."
   },
   "skills": {
     "label": "04 — SKILLS",

src/messages/pt.json

@@ -88,7 +88,8 @@
     "7_desc": "Real-time urban mobility analytics on TimescaleDB + PostGIS + Uber H3. Spring Boot ingestion, Kafka stream processing, Server-Sent Events for live dashboards. Smart-city back-end blueprint.",
     "8_desc": "Reference implementation of the BCBS 239 risk-data-aggregation lakehouse pattern on Databricks + Delta Lake + Unity Catalog + dbt-databricks. MIT-licensed, synthetic data only — banking risk-data engineering portfolio piece.",
     "9_desc": "Serverless middleware enforcing responsible-AI compliance on LLM applications — PII detection, bias screening, hate-speech filtering, and RAG-powered policy enforcement on top of Azure OpenAI.",
-    "10_desc": "Cloud Carbon Intelligence platform estimating CO2e emissions from Azure infrastructure with AI-powered reduction recommendations. Clean Architecture, Azure AI Search semantic factor lookup, 88 automated tests. Built for EU CSRD Scope 3 compliance."
+    "10_desc": "Cloud Carbon Intelligence platform estimating CO2e emissions from Azure infrastructure with AI-powered reduction recommendations. Clean Architecture, Azure AI Search semantic factor lookup, 88 automated tests. Built for EU CSRD Scope 3 compliance.",
+    "11_desc": "From-scratch ReAct Agent Observatory — observe, debug, and benchmark LLM agents with a built-in 8-type failure taxonomy (hallucinated tools, malformed actions, context overflow, goal drift, …) and multi-provider eval harness (Groq, OpenAI, Anthropic, Google, Ollama). Composite scoring (answer + tools + efficiency + reliability), Clean Architecture FastAPI + PostgreSQL backend, Next.js 16 frontend with real-time SSE streaming. 50+ benchmark cases, 81 tests."
   },
   "skills": {
     "label": "04 — SKILLS",

src/messages/th.json

@@ -88,7 +88,8 @@
     "7_desc": "Real-time urban mobility analytics on TimescaleDB + PostGIS + Uber H3. Spring Boot ingestion, Kafka stream processing, Server-Sent Events for live dashboards. Smart-city back-end blueprint.",
     "8_desc": "Reference implementation of the BCBS 239 risk-data-aggregation lakehouse pattern on Databricks + Delta Lake + Unity Catalog + dbt-databricks. MIT-licensed, synthetic data only — banking risk-data engineering portfolio piece.",
     "9_desc": "Serverless middleware enforcing responsible-AI compliance on LLM applications — PII detection, bias screening, hate-speech filtering, and RAG-powered policy enforcement on top of Azure OpenAI.",
-    "10_desc": "Cloud Carbon Intelligence platform estimating CO2e emissions from Azure infrastructure with AI-powered reduction recommendations. Clean Architecture, Azure AI Search semantic factor lookup, 88 automated tests. Built for EU CSRD Scope 3 compliance."
+    "10_desc": "Cloud Carbon Intelligence platform estimating CO2e emissions from Azure infrastructure with AI-powered reduction recommendations. Clean Architecture, Azure AI Search semantic factor lookup, 88 automated tests. Built for EU CSRD Scope 3 compliance.",
+    "11_desc": "From-scratch ReAct Agent Observatory — observe, debug, and benchmark LLM agents with a built-in 8-type failure taxonomy (hallucinated tools, malformed actions, context overflow, goal drift, …) and multi-provider eval harness (Groq, OpenAI, Anthropic, Google, Ollama). Composite scoring (answer + tools + efficiency + reliability), Clean Architecture FastAPI + PostgreSQL backend, Next.js 16 frontend with real-time SSE streaming. 50+ benchmark cases, 81 tests."
   },
   "skills": {
     "label": "04 — SKILLS",