[PR #333] docs: enable full DeepSeek V4 limits in Crush #351

New issue

Open

opened 2026-10-26 09:48:18 +00:00 by bsakaguchi · 0 comments

bsakaguchi commented

2026-10-26 09:48:18 +00:00

Owner

📋 Pull Request Information

Original PR: https://github.com/deepseek-ai/awesome-deepseek-agent/pull/333
Author: @Mr-Neutr0n
Created: 7/31/2026
Status: 🔄 Open

Base: main ← Head: docs/crush-v4-limits

📝 Commits (1)

2fed8bd docs: enable full DeepSeek V4 limits in Crush

📊 Changes

2 files changed (+28 additions, -12 deletions)

View changed files

📝 docs/crush.md (+14 -6)
📝 docs/crush.zh-CN.md (+14 -6)

📄 Description

Summary

replace the 1,048,576/32K model limits with DeepSeek V4's documented 1M context and 384K output
expose the max and high reasoning levels that Crush consumes and default both models to max
declare both V4 models text-only with supports_attachments: false
keep the English and Simplified Chinese guides synchronized

Current Crush source uses reasoning_levels to populate its reasoning selector and accepts default_reasoning_effort when it belongs to that list. Its own provider catalog already contains models with a 384K output limit and max reasoning, so these are active fields rather than inert documentation values.

Validation

parsed both JSON examples with jq
asserted both models in both guides have identical context, output, reasoning, and attachment values
checked current Crush schema and coordinator behavior for every added field
go test ./internal/config ./internal/agent against current Crush source
git diff --check

Checklist

Current V4 model names
1M context and 384K output configured
Maximum reasoning configured and selected by default
Text-only capability configured
English and Simplified Chinese updated together
No third-party workaround added

Open PR #282 also touches these files, but only to add a generic third-party disclaimer across every guide; it does not overlap these configuration changes.

Prepared with OpenAI Codex assistance; I reviewed the primary sources and changes.

_{🔄 This issue represents a GitHub Pull Request. It cannot be merged through Gitea due to API limitations.}

## 📋 Pull Request Information **Original PR:** https://github.com/deepseek-ai/awesome-deepseek-agent/pull/333 **Author:** [@Mr-Neutr0n](https://github.com/Mr-Neutr0n) **Created:** 7/31/2026 **Status:** 🔄 Open **Base:** `main` ← **Head:** `docs/crush-v4-limits` --- ### 📝 Commits (1) - [`2fed8bd`](https://github.com/deepseek-ai/awesome-deepseek-agent/commit/2fed8bd05a0044aa374efc2f38a47016eb6d2445) docs: enable full DeepSeek V4 limits in Crush ### 📊 Changes **2 files changed** (+28 additions, -12 deletions) <details> <summary>View changed files</summary> 📝 `docs/crush.md` (+14 -6) 📝 `docs/crush.zh-CN.md` (+14 -6) </details> ### 📄 Description ## Summary - replace the 1,048,576/32K model limits with DeepSeek V4's documented 1M context and 384K output - expose the `max` and `high` reasoning levels that Crush consumes and default both models to `max` - declare both V4 models text-only with `supports_attachments: false` - keep the English and Simplified Chinese guides synchronized Current Crush source uses `reasoning_levels` to populate its reasoning selector and accepts `default_reasoning_effort` when it belongs to that list. Its own provider catalog already contains models with a 384K output limit and `max` reasoning, so these are active fields rather than inert documentation values. ## Validation - parsed both JSON examples with `jq` - asserted both models in both guides have identical context, output, reasoning, and attachment values - checked current Crush schema and coordinator behavior for every added field - `go test ./internal/config ./internal/agent` against current Crush source - `git diff --check` ## Checklist - [x] Current V4 model names - [x] 1M context and 384K output configured - [x] Maximum reasoning configured and selected by default - [x] Text-only capability configured - [x] English and Simplified Chinese updated together - [x] No third-party workaround added Open PR #282 also touches these files, but only to add a generic third-party disclaimer across every guide; it does not overlap these configuration changes. Prepared with OpenAI Codex assistance; I reviewed the primary sources and changes. --- <sub>🔄 This issue represents a GitHub Pull Request. It cannot be merged through Gitea due to API limitations.</sub>

bsakaguchi added the

pull-request

label

2026-10-26 09:48:18 +00:00

bsakaguchi referenced this issue

2026-10-26 09:48:27 +00:00

[PR #351] docs: add Gliding Horse integration guide (EN + zh-CN) #363