anomalyco/opencode
1
0
Fork 0
mirror of https://github.com/anomalyco/opencode.git synced 2026-07-22 02:05:26 -04:00
Code Issues 5.9k Packages Projects Releases 19 Wiki Activity

Out of ideas. Simple hello takes 10k tokens on blank new project with no files in directory #6537

New Issue
Closed
opened 2026-02-16 18:04:31 -05:00 by yindo · 6 comments
yindo commented 2026-02-16 18:04:31 -05:00
Owner
Copy Link
Copy Source

Originally created by @mhrf on GitHub (Jan 16, 2026).

Originally assigned to: @rekram1-node on GitHub.

Description

I installed opecode today. Did some oneshot game in other dir.
Then I noticed each request is taking ~10 000 tokens without any tool calls.

So I created new dir. No files. No AGENTS.md.
Created new single session: Hello there !. These are stats:

...
{
"id": "...",
"sessionID": "ses_...",
"messageID": "msg_...",
"type": "step-finish",
"reason": "stop",
"cost": 0.006835,
"tokens": {
"input": 11408,
"output": 12,
"reasoning": 365,
"cache": {
"read": 0,
"write": 0
}
}
}

I am confused. Please can someone direct me or help me to understand what is included in ~11 000 input tokens ?

How can I see RAW Unaltered request going to Google API server ?

Plugins

No response

OpenCode version

1.1.20

Steps to reproduce

Using Gemini 3 Flash Preview (I think its not related)

  1. Create new dir
  2. run opencode
  3. /new Hello there !
  4. exit
  5. opencode export > export.json
  6. select given single conversation
  7. read content - no signs why input tokens would be ~10 000

Screenshot and/or share link

No response

Operating System

No response

Terminal

No response

Originally created by @mhrf on GitHub (Jan 16, 2026). Originally assigned to: @rekram1-node on GitHub. ### Description I installed opecode today. Did some oneshot game in other dir. Then I noticed each request is taking ~10 000 tokens without any tool calls. So I created new dir. No files. No AGENTS.md. Created new single session: Hello there !. These are stats: ... { "id": "...", "sessionID": "ses_...", "messageID": "msg_...", "type": "step-finish", "reason": "stop", "cost": 0.006835, "tokens": { "input": 11408, "output": 12, "reasoning": 365, "cache": { "read": 0, "write": 0 } } } I am confused. Please can someone direct me or help me to understand what is included in ~11 000 input tokens ? How can I see RAW Unaltered request going to Google API server ? ### Plugins _No response_ ### OpenCode version 1.1.20 ### Steps to reproduce Using Gemini 3 Flash Preview (I think its not related) 1. Create new dir 2. run opencode 3. /new Hello there ! 4. exit 5. opencode export > export.json 6. select given single conversation 7. read content - no signs why input tokens would be ~10 000 ### Screenshot and/or share link _No response_ ### Operating System _No response_ ### Terminal _No response_
yindo closed this issue 2026-02-16 18:04:31 -05:00
yindo commented 2026-02-16 18:04:31 -05:00
Author
Owner
Copy Link
Copy Source

@github-actions[bot] commented on GitHub (Jan 16, 2026):

This issue might be a duplicate of existing issues. Please check:

  • #254: Ridiculously big usage of input tokens for simple tasks (nearly identical - simple input consumed 10.2k tokens)
  • #5951: Token count is way off (massive discrepancy in token tracking)
  • #6728: High Token Consumption Observed with Opus 4.5 in OpenCode (same excessive token usage pattern)

Additionally, these closely related issues may explain the high token count:

  • #8887: Snapshot module ignores 'watcher.ignore' config, causing excessive disk usage (may add hidden context even in blank dirs)
  • #3182: Very slow session due to 98GB of git snapshots (snapshots adding massive context)
  • #6845: HUGE snapshot folder (snapshot accumulation issue)
  • #4113: Is there any way to debug compaction/token usage/prompts? (compaction adding 40k tokens)

Root cause analysis suggests the issue is likely related to the snapshot system adding unwanted context or fundamental token counting issues.

Feel free to ignore if none of these address your specific case.

@github-actions[bot] commented on GitHub (Jan 16, 2026): This issue might be a duplicate of existing issues. Please check: - #254: Ridiculously big usage of input tokens for simple tasks (nearly identical - simple input consumed 10.2k tokens) - #5951: Token count is way off (massive discrepancy in token tracking) - #6728: High Token Consumption Observed with Opus 4.5 in OpenCode (same excessive token usage pattern) Additionally, these closely related issues may explain the high token count: - #8887: Snapshot module ignores 'watcher.ignore' config, causing excessive disk usage (may add hidden context even in blank dirs) - #3182: Very slow session due to 98GB of git snapshots (snapshots adding massive context) - #6845: HUGE snapshot folder (snapshot accumulation issue) - #4113: Is there any way to debug compaction/token usage/prompts? (compaction adding 40k tokens) Root cause analysis suggests the issue is likely related to the snapshot system adding unwanted context or fundamental token counting issues. Feel free to ignore if none of these address your specific case.
yindo commented 2026-02-16 18:04:31 -05:00
Author
Owner
Copy Link
Copy Source

@rekram1-node commented on GitHub (Jan 16, 2026):

  • System prompt (including AGENTS.md files if u have them)
  • Tools & their prompts (including mcp servers if any)

You can actually view all the raw stuff in the Desktop app, we are going to add it to tui soon too.

If you wanna look at http stuff outside of apps I just setup a network proxy: https://opencode.ai/docs/network/

And monitor requests

@rekram1-node commented on GitHub (Jan 16, 2026): - System prompt (including AGENTS.md files if u have them) - Tools & their prompts (including mcp servers if any) You can actually view all the raw stuff in the Desktop app, we are going to add it to tui soon too. If you wanna look at http stuff outside of apps I just setup a network proxy: https://opencode.ai/docs/network/ And monitor requests
yindo commented 2026-02-16 18:04:31 -05:00
Author
Owner
Copy Link
Copy Source

@mhrf commented on GitHub (Jan 16, 2026):

  • System prompt (including AGENTS.md files if u have them)
  • Tools & their prompts (including mcp servers if any)

You can actually view all the raw stuff in the Desktop app, we are going to add it to tui soon too.

If you wanna look at http stuff outside of apps I just setup a network proxy: https://opencode.ai/docs/network/

And monitor requests

I see, is there a plan to edit BUILD-IN system prompt + SKILLs + MCP servers MD files ?
10k feels too much (I may not need all the buildin features).

@mhrf commented on GitHub (Jan 16, 2026): > * System prompt (including AGENTS.md files if u have them) > * Tools & their prompts (including mcp servers if any) > > You can actually view all the raw stuff in the Desktop app, we are going to add it to tui soon too. > > If you wanna look at http stuff outside of apps I just setup a network proxy: https://opencode.ai/docs/network/ > > And monitor requests I see, is there a plan to edit **BUILD-IN** system prompt + SKILLs + MCP servers MD files ? 10k feels too much (I may not need all the buildin features).
yindo commented 2026-02-16 18:04:31 -05:00
Author
Owner
Copy Link
Copy Source

@rekram1-node commented on GitHub (Jan 16, 2026):

@mhrf there arent built in skills, you can completely edit and configure all of this already on ur end.

You can override the agent prompts, you can disable tools, you can make completely new custom tools.

All of it can be controlled

@rekram1-node commented on GitHub (Jan 16, 2026): @mhrf there arent built in skills, you can completely edit and configure all of this already on ur end. You can override the agent prompts, you can disable tools, you can make completely new custom tools. All of it can be controlled
yindo commented 2026-02-16 18:04:32 -05:00
Author
Owner
Copy Link
Copy Source

@mhrf commented on GitHub (Jan 16, 2026):

@mhrf there arent built in skills, you can completely edit and configure all of this already on ur end.

You can override the agent prompts, you can disable tools, you can make completely new custom tools.

All of it can be controlled

Thank you for your help, I will look into it and how can I adjust it to fit my needs.

@mhrf commented on GitHub (Jan 16, 2026): > [@mhrf](https://github.com/mhrf) there arent built in skills, you can completely edit and configure all of this already on ur end. > > You can override the agent prompts, you can disable tools, you can make completely new custom tools. > > All of it can be controlled Thank you for your help, I will look into it and how can I adjust it to fit my needs.
yindo commented 2026-02-16 18:04:32 -05:00
Author
Owner
Copy Link
Copy Source

@rekram1-node commented on GitHub (Jan 16, 2026):

Relevant:
https://opencode.ai/docs/agents/

https://opencode.ai/docs/tools/

https://opencode.ai/docs/config/

This should help you out

@rekram1-node commented on GitHub (Jan 16, 2026): Relevant: https://opencode.ai/docs/agents/ https://opencode.ai/docs/tools/ https://opencode.ai/docs/config/ This should help you out
Sign in to join this conversation.
No Branch/Tag Specified
Branches Tags
Labels
Clear labels
acp
bad
bug
discussion
docs
documentation
enhancement
extension
good first issue
growth
help-wanted
mcp
model-problem
needs:compliance
nix
opentui
perf
pull-request
Mirrored from GitHub Pull Request
 tui
web
windows
zen
No Label
Milestone
No items
No Milestone
Projects
Clear projects
No project
Assignees
Clear assignees
yindo
No Assignees
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: anomalyco/opencode#6537
Delete Branch "%!s()"

Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?