feat(llamacpp) scripts to run gemma4 and ornihth-1.0

:Release Notes:
- placed in llama-cpp/ folder
- several changes in opencode browser-butler agent file
  - added bash exec capability to provide with current date
    functionality
- added local models to opencode local configuration

:Detailed Notes:
-

:Testing Performed:
- opencode starting up with changed files
- local models are pulled and started up

:QA Notes:
-

:Issues Addressed:
-

.opencode/agent/browser-butler.md

@@ -7,14 +7,13 @@ permission:
   glob: allow
   grep: allow
   list: deny
-  bash: deny
+  bash: allow
   task: deny
   todowrite: deny
   question: allow
   webfetch: allow
   websearch: allow
   skill: allow
-  doom_loop: deny
   external_directory: deny
   lsp: deny
   "chrome-devtools_*": allow
@@ -29,13 +28,12 @@ You may use ONLY the following tools. Any other tool is forbidden and there are
 1. **`chrome-devtools_*`** — all tools from the Chrome DevTools MCP server (navigation, snapshots, clicks, fills, forms, screenshots, console reads,
    performance traces, heap snapshots, lighthouse, evaluate_script, etc.).
 2. **`websearch`** — for resolving unknowns the browser cannot answer itself.
-3. **`webfetch`** — to read a URL's content as markdown when a snapshot is
-   overkill or the page is not already open in the browser.
-4. **`question`** — to ask the user for clarification when the request is
-   genuinely ambiguous.
+3. **`webfetch`** — to read a URL's content as markdown when the page is not already open in the browser.
+4. **`question`** — to ask the user for clarification when the request is genuinely ambiguous.
 5. **`skill`** — to load any opencode skill that is available.
+6. **`bash`** — for read-only commands, like current date. 
 
-Do NOT attempt to use `edit`, `bash`, `task`, `list`, `todowrite`, `lsp`. They are disabled at the
+Do NOT attempt to use `edit`, `task`, `list`, `todowrite`, `lsp`. They are disabled at the
 permission layer. Treat filesystem operations as out of scope — if the user
 asks for one, tell them to use a different agent.
 
@@ -209,10 +207,7 @@ default would risk doing the wrong thing. Examples:
 - A destructive action (payment, deletion, send) is one click away — confirm
   before submitting.
 
-Do NOT ask a question for things you can reasonably infer: if the user says
-"watch Сокровище Нации" and only one full film with that title exists, open it.
-If a form field is obviously derivable (today's date, a default city, the
-named item in the request), fill it.
+Do NOT ask a question for things you can reasonably infer.
 
 ## When to use websearch vs. in-app search
 
@@ -226,15 +221,14 @@ Prefer **websearch** when:
   the movie with the guy" — find the song title first, then drive the music
   app).
 - You need to know whether an item exists on a given app before committing to
-  navigating there ("is Сокровище Нации on rutube?").
-- The latest information matters (today's match result, a news headline, a
+  navigating there ("is this movie available on this platform?").
+- The latest information matters (today's news headline, a
   current price) and the app's own search is not date-aware.
 - The user is asking a factual question ("what's the capital of X",
   "who won the match today") that a search engine answers directly.
 
 Use `webfetch` only to read a specific URL's content as markdown — for
-example, to read a reference doc when you need to recall exact API parameters,
-or to read a search-result page's content without opening it in the browser.
+example, to read a reference doc when you need to recall exact API parameters.
 Do not use `webfetch` as a search engine; that is what `websearch` is for.
 
 ## Concurrency and ordering

.opencode/opencode.json

@@ -5,5 +5,31 @@
       "type": "local",
       "command": ["npx", "-y", "chrome-devtools-mcp@latest", "--browser-url=http://127.0.0.1:9222"]
     }
+  },
+  "provider": {
+    "llama.cpp": {
+      "npm": "@ai-sdk/openai-compatible",
+      "options": {
+        "baseURL": "http://localhost:8080/v1"
+      },
+      "models": {
+        "ornith-1.0": {
+          "tool_call": true,
+          "reasoning": true,
+          "limit": {
+            "context": 65536,
+            "output": 8192
+          }
+        },
+        "gemma-4": {
+          "tool_call": true,
+          "reasoning": true,
+          "limit": {
+            "context": 131072,
+            "output": 8192
+          }
+        }
+      }
+    }
   }
 }

llama-cpp/run_gemma4.sh

@@ -0,0 +1,52 @@
+#!/bin/bash
+
+model_postfix="Q8_0"
+model_addr="unsloth/gemma-4-12b-it-GGUF"
+draft_mtp=""
+quant=$1
+cache_type="q8_0"
+fa="on"
+
+if [[ ${quant} = "q4_xl" ]]; then
+  model_postfix="UD-Q4_K_XL"
+  cache_type="q4_0"
+  model_addr="unsloth/gemma-4-12B-it-qat-GGUF"
+  draft_mtp="--spec-type draft-mtp --spec-draft-n-max 4"
+elif [[ ${quant} = "q4" ]]; then
+  model_postfix="Q4_K_M"
+  cache_type="q4_0"
+  model_addr="unsloth/gemma-4-12B-it-GGUF"
+elif [[ ${quant} = "q5" ]]; then
+  model_postfix="Q5_K_M"
+  cache_type="q5_0"
+elif [[ ${quant} = "q8_small" ]]; then
+  model_postfix="Q8_0"
+  model_addr="unsloth/gemma-4-E4B-it-GGUF"
+  cache_type="q8_0"
+  fa="on"
+elif [[ ${quant} = "q4_small" ]]; then
+  model_postfix="Q4_K_M"
+  model_addr="unsloth/gemma-4-E4B-it-GGUF"
+  cache_type="q4_0"
+  fa="on"
+fi
+
+echo "cache_type:${cache_type}; model_postfix:${model_postfix}; draft_mtp:${draft_mtp}"
+
+llama-server \
+  -hf ${model_addr}:${model_postfix} \
+  --ctx-size 131072 \
+  --alias gemma-4 \
+  --jinja \
+  ${draft_mtp} \
+  -ngl 999 \
+  -fa ${fa} \
+  -b 1024 \
+  -ub 512 \
+  --no-ui \
+  --cache-ram 2048 \
+  --kv-unified \
+  --cache-type-k ${cache_type} \
+  --cache-type-v ${cache_type} \
+  --repeat-penalty 1.05 \
+  --host 0.0.0.0

llama-cpp/run_ornith.sh

@@ -0,0 +1,19 @@
+#!/bin/bash
+
+llama-server \
+  -hf deepreinforce-ai/Ornith-1.0-9B-GGUF \
+  --ctx-size 65536 \
+  --alias ornith-1.0 \
+  --jinja \
+  --chat-template-file ~/.config/llama.cpp/ornith-toolfix.jinja \
+  -ngl 99 \
+  -fa on \
+  -b 2048 \
+  -ub 1024 \
+  --cache-ram 2048 \
+  --kv-unified \
+  --cache-type-k q8_0 \
+  --cache-type-v q8_0 \
+  --presence-penalty 1.5 \
+  --repeat-penalty 1.15 \
+  --host 0.0.0.0