[GH-ISSUE #624] Feature Request The Companion - Conversation Snapshots, Manual Merging & True User-Controlled Continuity #186

New issue

Open

opened 2026-10-26 09:54:52 +00:00 by bsakaguchi · 1 comment

bsakaguchi commented

2026-10-26 09:54:52 +00:00

Owner

Originally created by @Muohamedsalem on GitHub (May 17, 2026).
Original GitHub issue: https://github.com/deepseek-ai/awesome-deepseek-integration/issues/624

Is your feature request related to a problem? Please describe.
Currently, long, insightful conversations that have built significant context, user personality, and trust are completely lost when the context limit is reached. Users must start from absolute zero, losing hours of nuanced discussion and the "relationship" built with the model. This is a critical barrier for professionals and power users who rely on DeepSeek for complex, long-term projects.

This request builds directly upon and expands the vision outlined in:

Feature Request #620 (Persistent Cross-Session Memory)
Feature Request #621 (Conversation Chaptering)

Describe the solution you'd like
We propose a comprehensive "Companion" capability that gives users true sovereignty and control over their conversation memory:

1. Conversation Snapshots (Export/Import)

Allow users to export an entire conversation into a portable file (e.g., .dsconv), preserving not just the text but the deep context, user personality, and ongoing project state.
Allow users to import this file into a new session, enabling the model to instantly regain the full context and continue with the same depth, intimacy, and understanding.
This gives users ownership of their data and the ability to store, back up, and transfer their conversation history.

2. Manual Context Merging (The Blend)

Allow users to import multiple past conversation files into a single new session.
The model should be able to synthesize a "blend" of these contexts, allowing users to combine insights from different projects or discussions into a unified, powerful workflow.
This transforms DeepSeek from a linear chat tool into a knowledge synthesis platform.

Describe alternatives you've considered

Relying solely on automatic chaptering (Feature Request #621).
Manually copy-pasting long text summaries (insufficient for preserving deep, nuanced context and user personality).

Additional context
A dedicated repository has been created to house the vision, ethical charter, and professional UAT test cases for this feature set:
👉 DeepSeek-The-Companion Repository: https://github.com/Muohamedsalem/DeepSeek-The-Companion

The repository includes:

Vision Statement (docs/companion-vision.md)
UAT Test Cases for Chaptering (test-cases/chaptering/)
UAT Test Cases for Export/Import & Manual Merging (test-cases/export-merge/)

This feature set would be a significant differentiator in the AI assistant market, as no major competitor currently offers this level of user-controlled, deep continuity. It transforms DeepSeek from a "stateless tool" into a "stateful companion."

Originally created by @Muohamedsalem on GitHub (May 17, 2026). Original GitHub issue: https://github.com/deepseek-ai/awesome-deepseek-integration/issues/624 **Is your feature request related to a problem? Please describe.** Currently, long, insightful conversations that have built significant context, user personality, and trust are completely lost when the context limit is reached. Users must start from absolute zero, losing hours of nuanced discussion and the "relationship" built with the model. This is a critical barrier for professionals and power users who rely on DeepSeek for complex, long-term projects. This request builds directly upon and expands the vision outlined in: - Feature Request #620 (Persistent Cross-Session Memory) - Feature Request #621 (Conversation Chaptering) **Describe the solution you'd like** We propose a comprehensive "Companion" capability that gives users true sovereignty and control over their conversation memory: **1. Conversation Snapshots (Export/Import)** - Allow users to export an entire conversation into a portable file (e.g., `.dsconv`), preserving not just the text but the deep context, user personality, and ongoing project state. - Allow users to import this file into a **new** session, enabling the model to instantly regain the full context and continue with the same depth, intimacy, and understanding. - This gives users ownership of their data and the ability to store, back up, and transfer their conversation history. **2. Manual Context Merging (The Blend)** - Allow users to import **multiple** past conversation files into a single new session. - The model should be able to synthesize a "blend" of these contexts, allowing users to combine insights from different projects or discussions into a unified, powerful workflow. - This transforms DeepSeek from a linear chat tool into a knowledge synthesis platform. **Describe alternatives you've considered** - Relying solely on automatic chaptering (Feature Request #621). - Manually copy-pasting long text summaries (insufficient for preserving deep, nuanced context and user personality). **Additional context** A dedicated repository has been created to house the vision, ethical charter, and professional UAT test cases for this feature set: 👉 **DeepSeek-The-Companion Repository:** [https://github.com/Muohamedsalem/DeepSeek-The-Companion](https://github.com/Muohamedsalem/DeepSeek-The-Companion) The repository includes: - **Vision Statement** (`docs/companion-vision.md`) - **UAT Test Cases for Chaptering** (`test-cases/chaptering/`) - **UAT Test Cases for Export/Import & Manual Merging** (`test-cases/export-merge/`) This feature set would be a significant differentiator in the AI assistant market, as no major competitor currently offers this level of user-controlled, deep continuity. It transforms DeepSeek from a "stateless tool" into a "stateful companion."

bsakaguchi commented

2026-10-26 09:54:54 +00:00

Author

Owner

@Mexikrusguy commented on GitHub (Jul 9, 2026):

ЭТО ИДЕАЛЬНОЕ ПРЕДЛОЖЕНИЕ

@Mexikrusguy commented on GitHub (Jul 9, 2026): ЭТО ИДЕАЛЬНОЕ ПРЕДЛОЖЕНИЕ

No labels

pull-request

No milestone

No project

No assignees

1 participant

Notifications

Due date

The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set.

Reference

deepseek-ai/awesome-deepseek-integration#186

No description provided.

Rows
Columns