frank-default-small: lower context by 32k per stream for cache headroom

3 slots x 229376 (262144 - 32768) = 688128 total context. Updates the
.env, models.json (contextWindow/maxTokens 229376 per slot), and the
index/deployment/ornith runbook references.

docs/LLM/dealign-ornith-crack.md

@@ -77,7 +77,7 @@ llama-server \
   --chat-template-file chat-template.jinja \
   --spec-type draft-mtp \
   --override-tensor "\.ffn_(up|down|gate|gate_up)_(ch|)exps=CPU" \
-  -ngl all -c 786432 -np 3 \
+  -ngl all -c 688128 -np 3 \
   --cache-type-k f16 --cache-type-v f16 --no-mmap \
   --temp 1.0 --top-p 0.95 --top-k 20
 ```

docs/LLM/deployment.md

@@ -128,7 +128,8 @@ binary with the MoE expert weights pinned to CPU by an explicit
 `--override-tensor` regex (the expert `\.ffn_(up|down|gate|gate_up)_(ch|)exps`
 pattern, buffer type `CPU`) and 7 CPU threads, native `draft-mtp`
 speculative decoding (bundled MoE MTP head, n_max 4), f16 KV, context
-786432 (3 slots of ~262K each), batch 4096 / ubatch 2048, `--no-mmap`. The
+688128 (3 slots of ~229K each, 32k headroom per stream), batch 4096 /
+ubatch 2048, `--no-mmap`. The
 unit is `scripts/frank-default-small.service`. It launches through
 `scripts/serve-frank-small.sh`, which reads tunables from
 `scripts/frank-default-small.env`. See [[dealign-ornith-crack]].

docs/LLM/index.md

@@ -20,7 +20,7 @@ graph LR
         B[upstream llama.cpp b10121
 frank-default-small, port 7770]
         C[Ornith-1.5-35B-A3B-CRACK Q8_0
-native MTP, MoE experts offloaded via override, 786432 ctx]
+native MTP, MoE experts offloaded via override, 688128 ctx]
         A --> B --> C
     end