Fix slot selection logic in get_available_slot

The bug was likely introduced in PR ikawrakow#973 when the similarity calculation
was changed from LCP to token-level similarity, but sim_best was still
initialized to 0 instead of -1.0f.

When slot_prompt_similarity threshold was set high (e.g., 0.8) and no slot
met the threshold, sim_best stayed at 0, causing ret to remain nullptr.
This led to the system getting stuck without selecting any slot.

This fix:
- Changed sim_best initialization from 0 to -1.0f
- Added best_slot variable to track the best slot found during similarity search
- Only set ret = best_slot after the loop completes
- Removed redundant ret == nullptr check

This ensures that even when no slot meets the slot_prompt_similarity threshold,
the system still identifies the best available slot and falls back to LRU correctly.

Related: PR ikawrakow#973 (Server: Handle context shift better), PR ikawrakow#1285 (Fix slot prompt updating)

Author: FixBot

examples/server/server-context.cpp

@@ -690,9 +690,10 @@ server_slot* server_context::get_available_slot(const server_task& task) {
     bool update_cache = false;
 
     // find the slot that has at least n% prompt similarity
-    if (ret == nullptr && slot_prompt_similarity != 0.0f) {
+    if (slot_prompt_similarity != 0.0f) {
         int max_lcp_len = 0;
-        float sim_best = 0;
+        float sim_best = -1.0f;
+        server_slot* best_slot = nullptr;
 
         for (server_slot& slot : slots) {
             // skip the slot if it is not available
@@ -720,13 +721,14 @@ server_slot* server_context::get_available_slot(const server_task& task) {
             float sim_cur = sim.second;
 
             // select the current slot if the criteria match
-            if (sim_cur > sim_best && sim_cur > slot_prompt_similarity) {
+            if (sim_cur > slot_prompt_similarity && sim_cur > sim_best) {
                 sim_best = sim_cur;
                 max_lcp_len = lcp_len.first;
-                ret = &slot;
+                best_slot = &slot;
             }
         }
-        if (ret != nullptr) {
+        if (best_slot != nullptr) {
+            ret = best_slot;
             LOG_VERBOSE("selected slot by lcp similarity", {
                 {"id_slot", ret->id},
                 {"max_lcp_len", max_lcp_len},
@@ -749,7 +751,6 @@ server_slot* server_context::get_available_slot(const server_task& task) {
                 ret = &slot;
             }
         }
-
         if (ret != nullptr) {
             LOG_VERBOSE("selected slot by lru", {
                 {"id_slot", ret->id},
@@ -766,7 +767,6 @@ server_slot* server_context::get_available_slot(const server_task& task) {
             if (exclude_think) {
                 auto temp = tokens.get_text_tokens_exclude_think(ret->ctx, ret->params.think_tokens);
                 server_tokens cache_exclude_think = server_tokens(temp, false);
-
                 temp = task.tokens.get_text_tokens_exclude_think(ret->ctx, ret->params.think_tokens);
                 server_tokens prompt_exclude_think = server_tokens(temp, false);
 
@@ -799,15 +799,13 @@ server_slot* server_context::get_available_slot(const server_task& task) {
             LLAMA_LOG_INFO("updating prompt cache\n");
             // copy cache tokens
             copy_data_to_cached_prompt(tokens, *ret);
-
             ret->prompt_save(*prompt_cache);
             LLAMA_LOG_INFO("prompt cache save took %.2f ms\n", (ggml_time_us() - t_start) / 1000.0);
         }
         // has prompts saved earlier to load
         if (prompt_cache && !prompt_cache->states.empty()) {
             const int64_t t_start = ggml_time_us();
             copy_data_to_cached_prompt(tokens, *ret);
-
             ret->prompt_load(*prompt_cache, task.tokens);
             prompt_cache->update();
 
@@ -821,6 +819,7 @@ server_slot* server_context::get_available_slot(const server_task& task) {
     return ret;
 }
 
+
 bool server_context::launch_slot_with_task(server_slot& slot, server_task& task) {
     slot_params defaults;
     defaults.speculative = params_base.speculative;