feat(embedding): add createEmbedding method to LLamaCPP class

feat(embedding): add llama_n_embd and llama_get_embeddings methods to LLamaCPPFFI class

test(LLamaCPPTest.php): add testCreateEmbedding method to test createEmbedding function in LLamaCPP class

examples/embedding.php

@@ -0,0 +1,21 @@
+<?php
+
+require_once __DIR__ . '/../vendor/autoload.php';
+
+use Kambo\LLamaCPP\LLamaCPP;
+use Kambo\LLamaCPP\Context;
+use Kambo\LLamaCPP\Parameters\ModelParameters;
+use Kambo\LLamaCPP\Parameters\GenerationParameters;
+
+$template = "You are a programmer, write PHP class that will add two numbers and print the result. Stop at class end.";
+$context = Context::createWithParameter(
+    new ModelParameters(
+        modelPath:__DIR__ .'/models/ggjt-model.bin',
+        embedding: true,
+    )
+);
+$llama = new LLamaCPP($context);
+
+$embeddings = $llama->createEmbedding($template);
+
+var_dump($embeddings);

src/LLamaCPP.php

@@ -92,4 +92,24 @@ public function generateAll(string $prompt, ?GenerationParameters $generation =
 
         return implode('', $tokens);
     }
+
+    public function createEmbedding(string $text, int $noOfThreads = 10): array
+    {
+        $input  = $this->ffi->newArray('llama_token', strlen($text));
+        $nOfTok = $this->ffi->llama_tokenize($this->context->getCtx(), $text, $input, strlen($text), true);
+
+        for ($i = 0; $i < $nOfTok; $i++) {
+            $this->ffi->llama_eval($this->context->getCtx(), $input + $i, 1, $i, $noOfThreads);
+        }
+
+        $nCount    = $this->ffi->llama_n_embd($this->context->getCtx());
+        $embedding = $this->ffi->llama_get_embeddings($this->context->getCtx());
+
+        $embeddings = [];
+        for ($i = 0; $i < $nCount; $i++) {
+            $embeddings[] = $embedding[$i];
+        }
+
+        return $embeddings;
+    }
 }

src/Native/LLamaCPPFFI.php

@@ -214,4 +214,21 @@ public function llama_token_eos()
     {
         return $this->fii->llama_token_eos();
     }
+
+    public function llama_n_embd(CData $ctx): int
+    {
+        return $this->fii->llama_n_embd($ctx);
+    }
+
+    /**
+     * Get the embeddings for the input
+     *
+     * @param CData $ctx
+     *
+     * @return ?CData
+     */
+    public function llama_get_embeddings(CData $ctx): ?CData
+    {
+        return $this->fii->llama_get_embeddings($ctx);
+    }
 }

tests/LLamaCPPTest.php

@@ -103,4 +103,40 @@ public function testGenerateAll()
 
         $this->assertEquals('test', $result);
     }
+
+    public function testCreateEmbedding()
+    {
+        $this->contextMock->method('getCtx')
+            ->willReturn(FFI::new('int'));
+
+        $this->ffiMock->method('newArray')
+            ->willReturn(FFI::new('int[100]'));
+
+        $this->ffiMock->method('llama_tokenize')
+            ->willReturn(1);
+
+        $this->ffiMock->method('llama_eval');
+
+        $this->ffiMock->method('llama_token_eos')
+            ->willReturn(0);
+
+        $this->ffiMock->method('llama_n_embd')
+            ->willReturn(5);
+
+        $testArray = FFI::new('int[5]');
+        foreach ([5,2,3,4,1] as $key => $value) {
+            $testArray[$key] = $value;
+        }
+        $this->ffiMock->method('llama_get_embeddings')
+            ->willReturn($testArray);
+
+        $llamaCPP = new LLamaCPP($this->contextMock, $this->eventDispatcherMock, $this->ffiMock);
+
+        $result = $llamaCPP->createEmbedding('test');
+
+        $this->assertEquals(
+            [5,2,3,4,1],
+            $result
+        );
+    }
 }