[BUG] Chat history lost after Antigravity update - ChatSessionStore index reset to empty

**Environment**

- Antigravity: latest version (updated 2026-02-22)

- OS: macOS Sequoia (Darwin 25.3.0)

- Hardware: MacBook Pro, Apple M4 Pro, 48GB RAM

- Plan: Google AI Pro

**What happened**

After updating Antigravity to the latest version, the majority of my chat history disappeared. Only sessions older than approximately one month are still visible. All recent sessions from the past several weeks are gone.

**Steps to reproduce**

1. Had multiple active chat sessions across 3 workspaces

2. Updated Antigravity to the latest version

3. After restart, most chat sessions were missing from the sidebar

**Technical investigation**

I examined the local database files and found:

- `chat.ChatSessionStore.index` in `state.vscdb` has been reset to `{ā€œversionā€:1,ā€œentriesā€:{}}` (completely empty)

- The backup database (`state.vscdb.backup`) also contains the same empty index (backup was overwritten)

- However, protobuf data in `jetskiStateSync.agentManagerInitState` still retains references to the lost sessions (UUIDs and titles), confirming the sessions existed before the update

**Lost sessions (titles recovered from protobuf)**

1. Demo URL Request

2. Refactoring To Monolith

3. Implementing Event System

4. Reset Project From Scratch

5. Fixing Upload Errors

6. Refining UI and Implementing Features

7. Understanding Chat Capabilities

8. Troubleshooting Indexing Errors

9. Setting Up RAG App

10. Enhancing Gemini RAG Output

**Affected workspaces**: RAGsystem demo, Glitch Protocol, ćƒ‘ćƒćƒƒćƒˆē”ØAI

**44 session UUIDs** still exist in `trajectorySummaries`, confirming sessions were not manually deleted.

**Request**

1. Is recovery possible? The UUIDs and session metadata are intact in protobuf state — conversation content may still exist server-side.

2. What caused `ChatSessionStore.index` to reset during the update? This appears to be a migration bug.

3. Is there a recommended backup method before future updates?

I have preserved the local database files (`state.vscdb`, protobuf binaries) and can provide them upon request.

16 Likes

Im experiencing the same thing, and it wasn’t even after an update. All my history is gone

5 Likes

I have the same problem

2 Likes

This worked for me! If you open the Agent Manager, you should be able to access your workspace history!

1 Like
4 Likes

Thank you for bringing the chat history issue to our attention. Our engineering team is currently investigating the matter, and we appreciate your patience as we work toward a resolution.

3 Likes

I’m having the same issue. It’s only showing conversations from a month ago, but my recent ones have disappeared.

1 Like

I’m having the same issue.

i have 8 powershell windows pop up anytime i try to open the Agent Manager and 3 pop up when i first launch the IDE. Plus when trying to reference my chat history some are there and some the agent just spins and spins trying to load (burning through tokens) I’ve spent past 2 hours trying to get Antigravity to investigate and help me fix the issue but i’m just burning through tokens to no end. Why not alllow us to revert back to the previous version while you guys work out the bugs.. Not being able to find chat history is a big deal.

1 Like

they have new version 1.19.4 but this bug still here. :frowning:

2 Likes

After I upgraded to 1.19.5, my recent chat history is cleared. However, an agent was working when I triggered an update and now it’s still working on background. I just cannot see the related conversation anywhere. Not in manager, not in editor. Since I have "Implementation Plan -tab open, I can command the agent by reviewing the implementation plan. Not even restarting the editor app help :sweat_smile:

Antigravity Version: 1.19.5
VSCode OSS Version: 1.107.0
Commit: 6adfc1a7e4a1a9af62bc45e8f2d7e6a97b7a9756
Date: 2026-02-26T07:23:14.771Z
Electron: 39.2.3
Chromium: 142.0.7444.175
Node.js: 22.21.1
V8: 14.2.231.21-electron.0
OS: Darwin arm64 25.2.0
Language Server CL: 875527528

This works for me! Steps to follow:

  • Open ā€œOpen Agent Managerā€
  • Click on ā€œOpen Workspaceā€
  • Open the folder you were working with
  • The chats history will appear
3 Likes

This also worked for me (now on 1.19.5).

This only happens to me with the Intel Mac version. The problem is that the FD service, which reads the files, is compiled for the Silicon Mac system, so it can’t execute the command and find the history files. The Intel Mac version needs all its services compiled for this platform.

1 Like

Same here. And we have ā€œImproved UI for banned usersā€ in the latest update, lol. Come on Google devs, the basic functionality doesn’t work, you know?

1 Like

Thanks a lot for sharing, worked here.

  • Open ā€œAgent Managerā€œ
  • ā€œFileā€ → ā€œOpen workspaceā€œ
  • Select your project folder (NOT THE WORKSPACE FILE)
    And your history will be there :100:

Step 1 — Rebuild the index (run while Antigravity is open, with at least 1 conversation visible): Save as rebuild_index.pyand run withpython rebuild_index.py:

import sqlite3, base64, os, re

DB = os.path.expandvars(r"%APPDATA%\antigravity\User\globalStorage\state.vscdb")
CONVS = os.path.expandvars(r"%USERPROFILE%\.gemini\antigravity\conversations")

conn = sqlite3.connect(DB)
cur = conn.cursor()
cur.execute("SELECT value FROM ItemTable WHERE key='antigravityUnifiedStateSync.trajectorySummaries'")
val = cur.fetchone()
conn.close()

if not val:
    print("ERROR: Start one conversation in a workspace first."); exit(1)

decoded = base64.b64decode(val[0])
current_uuid = re.findall(rb'[0-9a-f]{8}-[0-9a-f]{4}-[0-9a-f]{4}-[0-9a-f]{4}-[0-9a-f]{12}', decoded)[0].decode()

def rv(d,p):
    r,s=0,0
    while p<len(d):
        b=d[p];r|=(b&0x7F)<<s
        if(b&0x80)==0:return r,p+1
        s+=7;p+=1
    return r,p
def wv(v):
    r=b""
    while v>0x7F:r+=bytes([(v&0x7F)|0x80]);v>>=7
    r+=bytes([v&0x7F]);return r or b'\x00'

p=0;_,p=rv(decoded,p);l,p=rv(decoded,p);entry=decoded[p:p+l]

result,count=decoded,0
for f in sorted(os.listdir(CONVS)):
    if not f.endswith('.pb'):continue
    cid=f[:-3]
    if cid==current_uuid:continue
    cloned=entry.replace(current_uuid.encode(),cid.encode())
    result+=wv(0x0a)+wv(len(cloned))+cloned;count+=1

conn=sqlite3.connect(DB)
conn.cursor().execute("UPDATE ItemTable SET value=? WHERE key='antigravityUnifiedStateSync.trajectorySummaries'",(base64.b64encode(result).decode(),))
conn.commit();conn.close()
print(f"Injected {count} conversations. REBOOT YOUR PC (not just app restart).")

Then reboot your PC (full reboot, not just closing the app).

Step 2 — Fix titles (run AFTER closing Antigravity completely): All conversations will show the same placeholder title. To fix, save as fix_titles.py, close Antigravity, then run from cmd/PowerShell:

import sqlite3, base64, os, time

DB = os.path.expandvars(r"%APPDATA%\antigravity\User\globalStorage\state.vscdb")
BRAIN = os.path.expandvars(r"%USERPROFILE%\.gemini\antigravity\brain")
CONVS = os.path.expandvars(r"%USERPROFILE%\.gemini\antigravity\conversations")

titles = {}
for cid in (f[:-3] for f in os.listdir(CONVS) if f.endswith('.pb')):
    bp = os.path.join(BRAIN, cid)
    if not os.path.exists(bp): continue
    for item in os.listdir(bp):
        if item.startswith('.') or not item.endswith('.md'): continue
        with open(os.path.join(bp, item), 'r', encoding='utf-8', errors='replace') as f:
            line = f.readline().strip()
        if line.startswith('#'): titles[cid] = line.lstrip('# ')[:80]
        break

def ev(v):
    r=b""
    while v>0x7F:r+=bytes([(v&0x7F)|0x80]);v>>=7
    r+=bytes([v&0x7F]);return r or b'\x00'
def dv(d,p):
    r,s=0,0
    while p<len(d):
        b=d[p];r|=(b&0x7F)<<s
        if(b&0x80)==0:return r,p+1
        s+=7;p+=1
    return r,p
def esf(fn,s):
    b=s.encode('utf-8');return ev((fn<<3)|2)+ev(len(b))+b
def ebf(fn,b):
    return ev((fn<<3)|2)+ev(len(b))+b

conn=sqlite3.connect(DB);cur=conn.cursor()
cur.execute("SELECT value FROM ItemTable WHERE key='antigravityUnifiedStateSync.trajectorySummaries'")
decoded=base64.b64decode(cur.fetchone()[0])

entries,p=[],0
while p<len(decoded):
    t,np=dv(decoded,p)
    if(t>>3)!=1 or(t&7)!=2:break
    l,np=dv(decoded,np);entries.append(decoded[np:np+l]);p=np+l

rebuilt=b""
for entry in entries:
    ep,uid,ib=0,None,None
    while ep<len(entry):
        t,ep=dv(entry,ep)
        if(t&7)==2:
            l,ep=dv(entry,ep);c=entry[ep:ep+l];ep+=l
            if(t>>3)==1:uid=c.decode()
            elif(t>>3)==2:
                ip=0;it,ip=dv(c,ip);il,ip=dv(c,ip);ib=c[ip:ip+il].decode()
        elif(t&7)==0:_,ep=dv(entry,ep)
    if uid and ib:
        title=titles.get(uid)
        if not title:
            cp=os.path.join(CONVS,f"{uid}.pb")
            d=time.strftime("%b %d",time.localtime(os.path.getmtime(cp))) if os.path.exists(cp) else "?"
            title=f"Conversation ({d}) {uid[:8]}"
        idata=base64.b64decode(ib);fields,fp=[],0
        while fp<len(idata):
            sp=fp;t,fp=dv(idata,fp);fn,wt=t>>3,t&7
            if wt==0:_,fp=dv(idata,fp)
            elif wt==2:l,fp=dv(idata,fp);fp+=l
            elif wt==1:fp+=8
            elif wt==5:fp+=4
            else:break
            fields.append((fn,wt,idata[sp:fp]))
        nr=b""
        for fn,wt,raw in fields:
            nr+=esf(1,title) if fn==1 and wt==2 else raw
        ne=esf(1,uid)+ebf(2,esf(1,base64.b64encode(nr).decode()))
        rebuilt+=ebf(1,ne)
    else:rebuilt+=ebf(1,entry)

cur.execute("UPDATE ItemTable SET value=? WHERE key='antigravityUnifiedStateSync.trajectorySummaries'",(base64.b64encode(rebuilt).decode(),))
conn.commit();conn.close()
print(f"Fixed {len(titles)} real titles. Open Antigravity now.")

Important notes:
-
Step 1 must run while Antigravity IS open (needs an existing entry to clone)

  • Step 2 must run while Antigravity is CLOSED (otherwise the LS overwrites changes)
  • Step 1 requires a full PC reboot, not just an app restart
  • Requires Python 3 in PATH
  • Prevention: Always open Antigravity with a workspace folder to avoid this in the future

Hope this helps!

3 Likes

have the same problem
not after update by the way

I tried to fix this for several days, and finally had to ask an agent to clear all conversation history. He located the SQL database Antigravity is using and ran a script. (We made a backup of the history beforehand). Only after that did it get fixed. I just can’t understand WHY this hasn’t been fixed yet, so frustrating

Issue still persist in v1.20.4

1 Like