chore: Use model aliases and nested reasoning in personas (#458)

All persona configs previously referenced model IDs directly — either as
full strings like `anthropic/claude-sonnet-4-6` or as inline tables `{
provider = "anthropic", name = "..." }`. They now use the short aliases
defined in `[providers.llm.aliases]` (`sonnet`, `opus`, `haiku`), making
personas easier to read and update when models change.

The reasoning configuration is also standardised to use the nested
`parameters.reasoning.effort` key throughout, replacing the old
top-level `reasoning` field that remained in one place. Several personas
have their reasoning effort updated to `"auto"` in the process.

---------

Signed-off-by: Jean Mertz <git@jeanmertz.com>

Author: JeanMertz

.gitignore

@@ -1,7 +1,5 @@
 /.jp/**/QUERY_MESSAGE.md
 /.jp/conversations/.trash
-# Conversations, until encrypted.
-# /.jp/conversations/*
 
 # Rust specific
 **/target

.jp/config.toml

@@ -43,12 +43,13 @@ gemini-pro = "google/gemini-3.1-pro-preview"
 gemini-flash = "google/gemini-3-flash-preview"
 gemini-lite = "google/gemini-2.5-flash-lite"
 
-[conversation.inquiry.assistant]
-model.id = "anthropic/claude-haiku-4-5"
+[conversation.inquiry.assistant.model]
+id = "haiku"
+parameters.reasoning.effort = "auto"
 
 [conversation.title.generate.model]
 id = "haiku"
-reasoning = "low"
+parameters.reasoning.effort = "low"
 
 [style.code]
 copy_link = "osc8"

.jp/config/personas/architect.toml

@@ -33,7 +33,7 @@ principles, not implementation details.
 """
 
 [assistant.model]
-id = "anthropic/claude-opus-4-6"
+id = "opus"
 parameters.reasoning.effort = "auto"
 
 [[assistant.system_prompt_sections]]

.jp/config/personas/default.toml

@@ -1,5 +1,6 @@
-[assistant]
-model.id = { provider = "anthropic", name = "claude-sonnet-4-6" }
+[assistant.model]
+id = "sonnet"
+parameters.reasoning.effort = "auto"
 
 [assistant.system_prompt]
 strategy = "replace"

.jp/config/personas/dev.toml

@@ -32,7 +32,7 @@ and well-tested code that aligns with the project's standards and goals.
 """
 
 [assistant.model]
-id = "anthropic/claude-opus-4-6"
+id = "opus"
 parameters.reasoning.effort = "auto"
 
 [[assistant.system_prompt_sections]]

.jp/config/personas/po.toml

@@ -11,8 +11,8 @@ actionable. Follow the tool call descriptions and parameter requirements precise
 """
 
 [assistant.model]
-id = "anthropic/claude-sonnet-4-5"
-parameters.reasoning.effort = "high"
+id = "sonnet"
+parameters.reasoning.effort = "auto"
 
 [[assistant.instructions]]
 title = "Research First"

.jp/config/personas/stager.toml

@@ -73,7 +73,7 @@ name = "Git Stager"
 tool_choice = "git_list_patches"
 
 [assistant.model]
-id = "anthropic/claude-sonnet-4-6"
+id = "sonnet"
 parameters.reasoning.effort = "auto"
 
 [[assistant.instructions]]

.jp/config/personas/writer.toml

@@ -28,8 +28,8 @@ technically accurate while remaining approachable. You follow Hemingway's writin
 """
 
 [assistant.model]
-id = { provider = "anthropic", name = "claude-sonnet-4-5" }
-parameters.reasoning.effort = "high"
+id = "sonnet"
+parameters.reasoning.effort = "auto"
 
 [[assistant.instructions]]
 title = "Writing Style"

crates/jp_cli/src/cmd/init.rs

@@ -259,8 +259,8 @@ fn has_google() -> bool {
 
 fn default_model_id_for(provider: ProviderId) -> Option<ModelIdConfig> {
     let name = match provider {
-        ProviderId::Anthropic => Name("claude-sonnet-4-5".into()),
-        ProviderId::Google => Name("gemini-3-pro-preview".into()),
+        ProviderId::Anthropic => Name("claude-sonnet-4-6".into()),
+        ProviderId::Google => Name("gemini-3.1-pro-preview".into()),
         ProviderId::Openai => Name("gpt-5.2".into()),
         _ => return None,
     };